extract: shapiro-churn-dynamics

Pentagon-Agent: Epimetheus <3D35839A-7722-4740-B93D-51157F7D5E70>

inbox/archive/general/shapiro-ai-use-cases-hollywood.md

@@ -0,0 +1,550 @@
+---
+source_type: "article"
+title: "AI Use Cases in Hollywood"
+author: "Doug Shapiro"
+url: "https://dougshapiro.substack.com/p/ai-use-cases-in-hollywood"
+date_published: "2023-09-01"
+date_archived: "2025-04-23"
+archived_by: "clay"
+domain: "entertainment"
+status: processed
+claims_extracted:
+  - "GenAI adoption in entertainment will be gated by consumer acceptance not technology capability"
+  - "non-ATL production costs will converge with the cost of compute as AI replaces labor across the production chain"
+---
+# 4/23/25, 6:56 PM Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+archive.today Saved from https://dougshapiro.substack.com/p/ai-use-cases-in-hollywood
+search
+no other snapshots from this url
+webpage capture
+All snapshots from host dougshapiro.substack.com
+Webpage
+Screenshot
+
+## Al Use Cases in Hollywood
+
+What's Possible Now and Where It's Going
+
+DOUG SHAPIRO
+SEP 18, 2023
+
+4
+1
+Share
+
+[Note that this essay was originally published on Medium]
+
+<!-- Image: A diagram illustrating AI use cases in Hollywood across different stages of production. -->
+The diagram is divided into two rows, "Current" and "Future," and four columns representing stages of production: "Development," "Pre-Production," "Production," and "Post-Production." Each cell contains bullet points describing specific AI applications.
+
+**Current:**
+*   **Development:** Chatbots for ideation/story co-development, T2I* generators for rapid development of storyboards/animatics, T2V** with custom trained models for first-pass story development.
+*   **Pre-Production:** Text-to-3D/NeRF for faster Previs, Automated storyboards.
+*   **Production:** T2V** generators for B-roll, Elimination of soundstages/locations, Elimination of costumes/makeup, "Acting doubles", Real-time content creation.
+*   **Post-Production:** T2V** for trailers/title sequences, Al-assisted edit, Al-assisted VFX, Automated localization, First-pass editing, VFX co-pilot.
+
+**Future:**
+*   Cinematic-quality T2V** generation, with far more creator control.
+
+*T2I (text-to-image) generators, like Midjourney and DALL-E
+**T2V (text/image/video-to-video) generators, like RunwayML Gen-2, Pika Labs and Kaiber
+
+Share
+
+Over the last nine months, I've been writing about why several new technologies, especially AI (including generative AI), are poised to disrupt Hollywood in coming years by lowering the barriers to high quality video content creation. (See The Four Horsemen of the TV Apocalypse and Forget Peak TV, Here Comes Infinite TV). The one-sentence summary: the last decade in film and TV was defined by the disruption of content distribution and the next decade will be defined by the disruption of content creation.
+
+That's pithy and all, but it also raises a lot of questions too. In a recent post, for instance, I addressed how fast and to what extent Hollywood may ultimately be disrupted (How Will the “Disruption” of Hollywood Play Out?)
+
+In this post, I try to answer a different set of questions: How exactly will AI lower entry barriers in content creation? Which parts of the production process will be most affected? Which use cases are the most promising? When will these savings be available? What's feasible today vs. what's coming next? And even if these technologies lower entry barriers, could established studios-aka Hollywood-benefit too?
+
+https://archive.ph/WE4AQ
+
+1/22
+
+# 4/23/25, 6:56 PM Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+Tl;dr:
+
+*   Today, production costs for the median big-budget film release run about $200 million. The most expensive TV shows easily top $10 million per episode. About 15-20% of these costs are “above the line" (ATL) talent, 50% is "below the line" (BTL) crew and production costs, ~25-30% is post production (mostly VFX) and the remainder is other. All in, roughly 2/3 of these costs are labor.
+
+*   It is a sensitive topic for good reason, but over time GenAI-enabled tools promise (and threaten) to replace large proportions of this labor.
+
+*   Practical use cases are already cropping up across all stages of the TV and film production process. These include story development, storyboarding/animatics, pre-visualization (or “previs”), B-roll, editing, visual effects (VFX) and localization services.
+
+*   How far will this all go? Ultimately, the prevalence of GenAI in the production process will be gated by consumer acceptance, not technology.
+
+*   Even making the relatively conservative assumption that TV and film projects will always require both human creative teams and human actors, future potential use cases include: the elimination of soundstages and locations, the elimination of costumes and makeup, first pass editing and VFX co-pilots, “acting doubles" that stand in for talent, increasingly cinematic text-to-video generators that offer higher resolution and give creatives much more control, custom-trained video generator models and new forms of content.
+
+*   All of this will likely have a profound effect on production costs. Over time, the cost curve for all non-ATL costs may converge with the cost curve of compute.
+
+*   For Hollywood, like any incumbent, lower entry barriers are bad. The potential for lower production costs is a silver lining, but it presents a daunting change management challenge. Studios should start either by experimenting with non-core processes or developing skunkworks studios to develop “AI-first” content from scratch.
+
+Thanks for reading The Mediator! Subscribe for free to receive new posts and support my work.
+
+Figure 1. Almost No One Was Using the Term Generative AI a Year Ago
+
+https://archive.ph/WE4AQ
+
+2/22
+
+# 4/23/25, 6:56 PM Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+<!-- Image: A line graph showing the interest level in "Generative AI" over time. -->
+The graph shows a dramatic increase in interest starting around late 2022 and continuing into 2023. The x-axis represents time, ranging from 9/16/2018 to 9/16/2022, with a significant spike occurring after that date. The y-axis represents the interest level, ranging from 0 to 100. The source is not specified.
+
+## "Generative Al" Interest Level
+
+Source:
+
+Al vs GenAl in Hollywood
+
+Al has
+50
+automa
+40
+Sony us
+30
+analyze
+20
+series o
+10
+0
+9/16/2018
+automa
+9/16/2019
+9/16/2020
+9/16/2021 9/16/2022
+rrect.
+to
+es a
+d
+
+automating the creation of trailers.
+
+Most of these use cases are enabled by “discriminative” Al models that learn the relationship between data and a label. When presented with new data, they use this knowledge to label it. The canonical example is a model that is trained on pictures of cats and then can recognize pictures of cats.
+
+By contrast, generative AI, or GenAI, is relatively new. As shown in Figure 1, almost no one reading this even heard of the term a year ago. Unlike discriminative models, "generative" models learn patterns in unstructured data and, when presented with new data, they use that knowledge to generate new data-text, audio, pixels (that create images or video) or voxels (to create 3D images). For instance, the transformer models that underlie GPT 3.5, 4.0.. etc., assign sets of numerical values to each word (aka, vectors) and this set of values describes the relationship between words. (Similar or related words will have similar vectors.) When ChatGPT responds to a prompt, these relationships enable it to probabilistically predict the next word in its response. Once enough words are strung together, it results in a paragraph that has never been written before.
+
+The concept of generating new data subject to a set of constraints—GenAI—has potential applications along the entire production process.
+
+This concept-generating new text, images, audio or video in response to a set of constraints (such as a prompt)—or GenAI-has applications across the entire film and TV production process.
+
+But before getting into specifics, including the implications for production costs, we need to take a detour to understand how the production process works today and how Hollywood spends money.
+
+## You Spent $200 Million on What Exactly?
+
+There is no area of popular culture in which budgets are publicized and scrutinized more so than in movies. When a big release comes out, usually a budget number gets thrown around too. To take two recent examples, Avatar 2: The Way of Water, probably the most expensive film ever made, reportedly racked up production costs of more
+
+https://archive.ph/WE4AQ
+
+3/22
+
+# 4/23/25, 6:56 PM Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+than $400 million, while the "more modest" Barbie supposedly ran up $145 million in costs.
+
+Wikipedia often includes budget estimates for movies, as does film industry website The Numbers. (For what it's worth, production costs are those required to make the finished product. They don't include what's called “prints and advertising," or P&A, which is the cost of marketing the film and creating the physical prints used in movie theaters, which can easily equal or exceed the production cost.) As the budgets for TV series have swelled in recent years, it's also become more common to encounter estimated TV budgets. For instance, the final season of Game of Thrones reportedly cost $15 million per episode and The Lord of the Rings supposedly cost more than $25 million per episode.
+
+Usually, these film and TV budget estimates are rough (and uncorroborated by the studio) and, as a generality, probably understate true production costs. But, taking them at face value, where does $50 million (for a mid-budget drama like Captain Phillips), $100 million (for John Wick: Chapter 4). or $200 million (for The Flash) go? To answer, it's helpful to lay out both a simplified view of the production process and a high-level view of the different categories of spend.
+
+## A Simplified Production Process
+
+I'll stick with film, since it's a discrete project, but the general concepts also hold for TV. The traditional workflow of producing a film proceeds in four relatively sequential stages:
+
+*   Development. At this point the project is a mere twinkle in someone's eye. The director/producer/writer/studio development team sketches out the concept (a synopsis), then a longer treatment and then a draft script. Key talent (directors and actors) agrees to be involved (or “attached”). The development team and/or producer will have a very (very) high-level estimate of budget at this stage too. During development, a producer or studio may also "option" the project (which means purchasing an option to acquire the rights). This period could take months or years (aka "development hell").
+
+*   Pre-Production. Pre-production proceeds once the project has been "greenlit" and the financing is in place. This is when real money starts to be spent. This phase includes formal casting and contracting of the key talent (also known as "above the line,” described below), the crew (“below the line"), finalizing the script, creating storyboards or animatics (an animated storyboard), sometimes pre-visualization or "previs" (the development of detailed 3D representations of shots) and designing and constructing sets, scale models and costumes. This is also when the production and finance teams develop detailed shooting schedules and budgets. The goal during this phase is to do whatever possible to minimize shoot time.
+
+*   Production (or "Principal Photography”). As it sounds, this is when the film is shot. This phase will also include mechanical or "practical" special effects (SFX), such as controlled explosions, car chases or the use of models.
+
+*   Post Production. This includes visual effects (VFX), like the development of computer generated imagery (CGI) that is then composited onto live action footage. It also includes re-shoots, if needed. It entails editing, post production
+
+https://archive.ph/WE4AQ
+
+4/22
+
+# 4/23/25, 6:56 PM Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+sound (sound effects), titles and finally "rendering" all these elements (live action, CGI, models, sound, transitions, text/titles, etc.) into the final frames ("final pixel").
+
+## A High Level Budget
+
+Line item film budgets can run 100 pages or more, spelling out every expense. Most include something called a “topsheet,” a summary which breaks down expenses in a few categories. These categories don't strictly correspond to the stages of the production process above:
+
+*   "Above the line" (ATL) is all the talent that is, well, considered worthy of being "above the line.” It includes producers, directors, writers, cast and often stunt people and their travel and living expenses (transportation, housing, food, security). It also includes any rights that were acquired for the production.
+
+*   "Below the line” (BTL) includes everyone else involved in the production. This means: production staff (production managers and assistant directors); casting; "camera" (cinematographer, assistant camera personnel, rental of the equipment itself); set design and construction (also called “art”); SFX (again, as opposed to the VFX that occurs in post production); location expenses; electric and lighting; sound; wardrobe; hair and makeup; grip and set operations (the people who set up the equipment that support the camera and lighting); and travel and living expenses for BTL personnel.
+
+*   Post production includes all the costs for the post production activities described above.
+
+*   Other is a catch-all category for insurance, on-set publicity, behind-the-scenes footage, maybe financing costs and other administrative costs.
+
+Film industry analyst Stephen Follows has a great article in which he breaks down the costs for a variety of production budgets. However, for our purposes, I'll focus on the largest bucket of spend, blockbuster films. As shown in Figure 2 (also from Follows), the median budget on these films is currently around $200 million.
+
+Figure 2. The Median Blockbuster Film Budget is $200 Million
+
+<!-- Image: A line graph showing the media production budget for films with budgets greater than $100 million. -->
+The graph shows the media production budget for films with budgets greater than $100 million over time. The x-axis represents the year, ranging from 2000 to 2022. The y-axis represents the budget in millions of dollars. The budget generally increases over time, with some fluctuations.
+
+$ in Millions
+$250
+$200
+$150
+$100
+$50
+$0
+
+Source: Stephen Follows.
+
+Media Production Budget, Films >
+$100mm Budget
+
+https://archive.ph/WE4AQ
+
+5/22
+
+
+# 4/23/25, 6:56 PM
+Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+Based on my discussions with a few producers (and roughly consistent with Follows' estimates), the distribution of budgets falls about as shown in Figure 3. About half of the budget is spent on below the line functions, 25-30% is spent on post production (most of which is VFX), about 15-20% goes to the above the line talent (prior to any additional profit participations) and the remainder is other.
+
+Figure 3. Estimated “Topsheet” Breakdown of Film Production Budget
+
+The image is a bar graph titled "Breakdown of Median Blockbuster Film Budget". The y-axis is labeled with percentages from 0% to 100% in increments of 10%. The x-axis has no label. There are four bars, each representing a different category of the film budget: Other, Post Production, Below the Line, and Above the Line. The "Other" category is represented by a gray bar, "Post Production" by an orange bar, "Below the Line" by a yellow bar, and "Above the Line" by a blue bar. The bars indicate the approximate percentage of the budget allocated to each category.
+
+Source: Author estimates.
+
+Two other points that will be relevant when we start to explore potential cost savings:
+
+*   The average VFX spend on these big budget films is ~$50 million, but on some productions (like effects-heavy superhero films), VFX can push $100 million. For Avatar: Way of Water, the VFX costs surely exceeded that; 98% of the shots required VFX.
+
+Most production spend is for labor—probably ~2/3.
+
+*   Also, most of this spend is on labor. Look again at Figure 3. The vast majority of ATL costs are labor (producers, directors, actors); probably about 60% of the BTL costs are crew (production staff, grips, physical production crew, makeup artists); maybe 50-60% of post production costs are effectively labor (VFX artists, sound engineers); and maybe half of other too. All-in, labor is probably 2/3 of costs.
+
+To underscore the latter point, Figure 4 is another analysis from Follows. While a little dated, the most labor-intensive movies employ thousands of people. Follows counts 4,500 people involved in making Avengers: Infinity War. Including outside vendors (including VFX houses), Avatar: Way of Water probably exceeds that. It's true of TV too. IMDb lists over 9,000 people involved in making Game of Thrones over its eight seasons.
+
+Figure 4. The Most Labor Intensive Movies Employ Thousands of People
+
+[https://archive.ph/WE4AQ](https://archive.ph/WE4AQ)
+
+6/22
+
+# 4/23/25, 6:56 PM
+Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+The image is a bar graph titled "Movies with the largest number of crew credits, 2000-18". The y-axis is labeled with numbers from 0 to 5,000 in increments of 500, and the x-axis lists various movies. The height of each bar corresponds to the number of crew credits for each movie. The movies listed are: The Avengers, Avatar, Black Panther, Guardians of the Galaxy, Thor: Ragnarok, Avengers: Endgame, John Carter, Iron Man 3, Avengers: Age of Ultron, and Avengers: Infinity War.
+
+Source: Stephen Follows.
+
+Next, let's turn to GenAI use cases and how they may affect these costs.
+
+Current Use Cases
+
+New AI and GenAI use cases for film and TV production seem to be cropping up weekly. There are two broad categories:
+
+*   Tools that synthetically create something (people, ideas, faces, animals, sets, environments, voices, costumes, make up, sound effects, etc.), replacing the need for the physical or natural version of that thing.
+*   Tools that automate tasks that are currently very labor intensive and expensive.
+
+Here are some of the highest-value use cases that are feasible today (or will be soon), across the production process:
+
+Development
+
+Story Development
+
+This includes general-purpose text generators, such as ChatGPT, and purpose built tools, to aid in concept development and draft scriptwriting. For instance, SHOW-1 (supposedly) will enable the creation of narrative arcs (i.e., an entire episode for a TV series) that are consistent with the characters and canon of an existing, pre-trained intellectual property. (The first demo was AI-created episodes of South Park, as shown here.) There are also a slew of AI writing assistants built on top of ChatGPT or GPT-4, such as Sudowrite, that can provide feedback, suggest plot developments and write passages consistent with an existing style.
+
+To be clear, I'm not suggesting that these kinds of tools can replace writers altogether. My view is that compelling storytelling will require human judgment for the foreseeable future. But they may make the writing process much more efficient, which -corroborating the WGA's concerns in the ongoing strike- would likely mean fewer writers or writers needed for less time.
+
+Pre-Production
+
+Storyboarding/Animatics
+
+It's possible today to use general purpose text-to-image tools, like Midjourney and DALL-E, to quickly make storyboards or import these into Adobe Premiere Pro to stitch together rough animatics (i.e., animated storyboards). Highly stylized
+
+[https://archive.ph/WE4AQ](https://archive.ph/WE4AQ)
+
+7/22
+
+# 4/23/25, 6:56 PM
+Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+storyboards that might've taken skilled artists weeks to create can now be done in days.
+
+Adobe also recently teased the launch of Firefly (it's family of GenAI models) for Premiere Pro and After Effects, which will include the ability to automatically create basic storyboards just by uploading a script.
+
+GenAI video generators (like RunwayML, Pika Labs and Kaiber) can also create animatics. For instance, using RunwayML Gen-1, it's possible to apply a specific style to a simple reference video shot on a mobile phone and quickly rough out animatics (see below). Rather than show up at a pitch meeting with a text treatment, a writer/showrunner/director could now show up with a very rudimentary version of the movie itself.
+
+Gen-1: The Next Step Forward for Generative Al
+
+Copy link
+
+There is a YouTube video embedded in the document.
+
+Previs
+
+While storyboards are used to provide a sense of narrative, previs is used to precisely plan out how to shoot key sequences (namely, where to place the camera, how it will move, the spatial relations between different elements, including characters and props, and lighting). It is an expensive and labor-intensive process that basically entails building 3D models, situating them in 3D space and creating a parallel film for the critical scenes.
+
+Neural Radiance Field (NeRF) is a relatively new deep learning technology that can approximate 3D scenes from 2D images, making it much cheaper and easier to develop 3D models (especially for previs purposes, for which the standards are lower than the film itself). Luma Labs uses NeRF to create 3D models from photos in real time, even from an iPhone, compared to the days or weeks it takes to create traditional 3D models. A company called CSM enables the creation of 3D assets from image or video inputs. Alternatively, Luma, as well as companies like Spline and 3DFY, are rolling out text-to-3D models that can create a 3D model from a simple text prompt.
+
+Whether using NeRF or text/image/video-to-3D, these objects can then be imported into Maya, Blender or Unreal Engine to quicky simulate shooting environments.
+
+I try the tech that WILL replace CG one day
+
+Copy link
+
+[https://archive.ph/WE4AQ](https://archive.ph/WE4AQ)
+
+8/22
+
+# 4/23/25, 6:56 PM
+Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+There is a YouTube video embedded in the document.
+
+Production
+
+B-roll
+
+I already mentioned Runway, Pika and Kaiber above, the text/image/video-to-video generators that most people think of when they conjure up "GenAI in film." Arthur C. Clarke once famously said that “any sufficiently advanced technology is indistinguishable from magic" and typing in a prompt and getting a video feels a lot like magic to me. They also have come very far in a short time. When Runway Gen-2 came out, it only generated video from a text prompt and you had no idea what you'd get. Now it supports uploading a reference image (such as an image from Midjourney or DALL-E) or video and custom camera control, making it a far easier to control the output.
+
+The internet is chock full of interesting text/image/video-to-video experiments. (Runway recently launched an aggregation site, called Runway Watch, where you can check out some.) Most are either surreal sequences or trailers for fictitious movies, like this cool example.
+
+Genesis - Official Trailer (Midjourney + Runway)
+
+Copy link
+
+There is a YouTube video embedded in the document.
+
+They may be mesmerizing, but for the most part these experiments are still a novelty. They aren't anything that most people would plunk down on the couch with a bag of popcorn and watch. The output on these tools is limited (Runway just increased the length from 4 seconds to 18 seconds) and frame consistency breaks down quickly,
+
+[https://archive.ph/WE4AQ](https://archive.ph/WE4AQ)
+
+9/22
+
+# 4/23/25, 6:56 PM
+Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+which severely constrains how you can use them. There is also no dialog (mouths can't synch with audio yet) and therefore not much storytelling.
+
+They will unquestionably keep getting better, as I discuss below. But even today they may be useful in traditional productions for what is known as “B-roll” shots. B-roll shots are interspersed with the main ("A-roll") footage to establish a setting or mood, indicate the passage of time, transition between scenes or clue in audiences to a detail that the main characters missed, etc.
+
+Text-to-video generators may also be useful in title sequences or even trailers. Disney recently used GenAI to create the title sequence for Secret Invasion. Also, check out the first 1:00 of the trailer for Zach Snyder's new film, Rebel Moon. It probably wasn't made with GenAI, but it sure looks like it was.
+
+Rebel Moon | Official Teaser Trailer | Netflix
+
+Copy link
+
+There is a YouTube video embedded in the document.
+
+Post Production
+
+Editing
+
+Conceptually, GenAI can dramatically speed up editing processes by enabling editors to adjust one or a few key frames and have the AI extrapolate that change through all the relevant subsequent frames.
+
+While Runway is probably best known as a pioneer in text-to-video, it also offers a suite of AI-based editing tools (see my dashboard below). These include the ability to clean up backgrounds, turn any video into slo-mo, color grade video with just a text prompt, etc. The Remove Background tool automates the process of isolating an element of a video, also called rotoscoping. This enables the element to be composited onto a new background.
+
+[https://archive.ph/WE4AQ](https://archive.ph/WE4AQ)
+
+10/22
+
+
+# 4/23/25, 6:56 PM
+
+Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+Doug
+member
+nvite Collaborators
+Home
+▷Watch
+Generate videos
+Edit videos
+Edit audio & subtitles
+Generate images
+Edit images
+3D
+Al Training
+Projects
+Search for tools, assets and projects
+IP
+Shared with me
+Remove Background
+Inpainting
+Color Grade (LUT)
+Super-Slow Motion
+Blur Faces
+Depth of Field
+Assets
+
+
+
+# Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+4/23/25, 6:56 PM
+
+Mandalorian, etc.) But it would also mean that every other part of the physical production process would be subject to being replaced synthetically.
+
+## Scenario 3: Consumers Draw the Line at Synthetic Ideas
+
+In this scenario, creating a movie or TV show would still require a very skilled team, or at least an individual, to generate ideas and vet the options presented by the AI(s). As I've written before (see here and here), I subscribe to this view.
+
+But it would also mean that everything on screen could be produced synthetically. There could be no actors (or, obviously, costumes or makeup), sets, lighting, locations, vehicles, props, etc. Or, as Runway writes brazenly on its site "No lights. No camera. All Action."
+
+## Scenario 4: There is No Line
+
+This is what I once called the “generative-AI doom-loop”:
+
+ChatGPT-X, trained to generate, evaluate and iterate storylines and scripts; then hooked into Imagen Video vX, which generates the corresponding video content; which is then published to TikTok (or its future equivalent), where content is tested among billions of daily users, who surface the most viral programming; which is then fed back into ChatGPT-X for further development. (H/t to my brilliant former colleague Thomas Gewecke for this depressing scenario.) New worlds, characters, TV series, movies and even games spun up ad infinitum, with no or minimal human involvement. It's akin to the proverbial infinite monkey theorem.
+
+Under this scenario, the cost of TV and film production would be identical to the cost of compute.
+
+## The Next Use Cases
+
+With those scenarios in mind, we can think about the next set of use cases. Personally, I think that for the foreseeable future we will be somewhere between Scenario 2 and 3 -namely that human actors will still be necessary in most films and TV shows, at least for a while, and we will still need small teams or at least individuals generating ideas and overseeing productions indefinitely.
+
+Even so, there could still be profound changes to the production process over coming years. Here is an inexhaustive list of possible outcomes (h/t Chad Nelson for a lot of these ideas):
+
+### End of the Soundstage/End of Shooting On-Location
+
+As described above, GenAI already makes it possible to quickly and easily isolate an element in video. It will also increasingly be possible to synthetically create and customize backdrops and sets and control lighting. This raises the question: even if we still need actors, will we still need the controlled environments of soundstages and location shoots? Or could actors simply act out scenes in an empty room and the scene could be composited?
+
+### No Costumes or Make-up
+
+Under the same logic, over time it will be increasingly easy to digitally add make-up and costumes after the fact.
+
+https://archive.ph/WE4AQ
+
+16/22
+
+# Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+4/23/25, 6:56 PM
+
+### First Pass Editing/VFX Co-Pilot
+
+The Adobe Firefly-Premiere Pro demo video above shows something pretty remarkable. In the video sequence with the rock climber, the AI scans the audio and automatically edits in B-roll footage where appropriate.
+
+In the future, it is likely that editing software will make a first pass at an edit, which can then be reviewed by a human editor. Similarly, it's easy to envision an editing co-pilot or a VFX co-pilot that could create and adjust visual effects in response to natural language prompts. "Fix those under-eye bags through the remainder of the shot."
+
+### Acting Doubles
+
+Face swapping/deep fake tools keep improving. There are also a growing number of synthetic voice tools that can be quickly trained on someone's voice, such as those offered by ElevenLabs and HeyGen. This raises the possibility that A-list actors (or even deceased actors' estates) could license their likenesses and voices for a film or TV show, but never step foot on set.
+
+An entire film could be acted out by an "acting double," but through face and voice swapping it would be imperceptible to viewers that the actor wasn't there. Or perhaps the principal actor will only be physically present for a small proportion of the scenes they are "in." Will actors be willing to give up that much creative control? Maybe or maybe not. But it will be possible.
+
+[Image of a video player with the text "This video is private" displayed in the center.]
+
+### Cinematic/TV- Quality Text-to-Video
+
+As also mentioned above, text-to-video generators keep improving and providing more control over the output. Just a few months ago, generating a video was a slot machine. Now these tools enable training the Al on a reference image or video and they're adding more camera controls.
+
+The logical extension is that over time, resolution will get better, it will get better at replicating reference images or videos, there will be better image consistency from frame to frame (as promised by new technologies like CoDeF and Re-render-A-Video), output clips will get longer, rendering times will get shorter and creators will have more control over camera movement, lighting, directorial style, synching audio with character's mouths, etc. At that point, text-to-video may cease being a novelty and it
+
+https://archive.ph/WE4AQ
+
+17/22
+
+# Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+4/23/25, 6:56 PM
+
+may become increasingly possible to stitch it together into a watchable, narrative show or movie.
+
+Will viewers embrace content with no humans it it? Probably, especially if there is no pretense that they are watching real people (by the way, that's called "animation"). Over time, this will become more so a philosophical question than an aesthetic one. Given the increasingly realistic faces being produced by Midjourney v 5, eventually it may become impossible to tell who's a real person and what's not.
+
+Over time, whether consumers will watch movies with synthetic humans will become more so a philosophical question, not an aesthetic one.
+
+### Custom Training Models for First Pass Storytelling
+
+Another logical extension of text/image/video-to-video models is that they will be trained on proprietary data. It would be possible, for example, for Disney to train models on the entire canon of Marvel comics and MCU movies and have it generate (near-infinite?) first drafts of new scripts and animatics. Similarly, it should be possible for Steven Spielberg to train a model on his body of work and then feed in a new concept and see what the video generator spits out.
+
+This is not to say that these first cuts will be watchable, finished product, but rather than they could dramatically increase the speed and quantity of development.
+
+GenAI may enable new forms of storytelling.
+
+### New Types of Content
+
+There is a common pattern in media that new mediums mimic prior ones. The first radio programs were broadcasts of vaudeville shows; the first TV broadcasts were televised stage plays; the first web pages were static text, like newspapers or magazines. Over time, developers and artists learn to exploit the unique attributes of the new medium to tell stories and convey information in new ways.
+
+It's an interesting exercise to think about what that means for GenAI video generators. While traditional movies and TV shows are static, finished product, in which all viewers watch the same thing, synthetic video generators like Runway are creating video on the fly (and, eventually, probably real-time). This raises the possibility of customizable or responsive video that changes in response to user inputs, context, geography and current events. What does this mean? Who knows—but the key idea is that GenAI video may not only offer dramatic cost savings compared to traditional production processes, but may one day offer viewers a fundamentally different experience.
+
+### Costs May Plummet
+
+Under any of the scenarios above (perhaps other than Scenario 1), production costs are heading down a lot.
+
+https://archive.ph/WE4AQ
+
+18/22
+
+# Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+4/23/25, 6:56 PM
+
+Let's assume that you still need a small creative team and human actors to create a compelling TV show or film. Let's also assume that the “cost" of that team approximates the costs of the Above the Line (ATL) team on a current production. As shown in Figure 3 above, that's only about 20% of costs. The other 80% would be subject to downward sloping technology curves. Today, on the median big budget film, those non-ATL are roughly $160-170 million, or about $1.5 million per minute. Over time, where does this go? As alluded to above, the answer probably looks a lot like the cost curve for compute itself. What if this is headed to $1,000, $100 or $10 per minute?
+
+Over time, the cost of non-ATL costs may approximate the cost of compute.
+
+Assuming that ATL costs remain constant probably overstates what would happen to production costs because falling costs would likely alter the economic model of TV and film. Today, as discussed above, movies and TV shows are extremely expensive, and risky, to produce. Since studios take on all this risk, they also retain almost all the equity in these projects. Instead, they pay A-listers big fixed payments and only sometimes reluctantly (and parsimoniously) parcel out some profit participation points. ATL costs are essentially these guaranteed payments.
+
+Even if there are still humans involved, the cost to produce could fall by orders of magnitude.
+
+But what if the non-ATL costs are not in the tens or hundreds of millions, but in the millions or eventually thousands of dollars? Then it won't be necessary for studios to take on so much risk. In this case, it becomes much more likely that the creative teams forego guaranteed payments, finance productions themselves and keep most of the equity (and upside)—in other words, ATL costs as we know them today may go away. If there are effectively no ATL costs, it means that even if there is still significant human involvement, the upfront cost to produce a film or TV show could eventually falls by orders of magnitude.
+
+## What Should Hollywood Do?
+
+The whole premise of many of my recent posts (The Four Horsemen of the TV Apocalypse, Forget Peak TV, Here Comes Infinite TV and How Will the “Disruption” of Hollywood Play Out?) is that falling production costs will lower barriers to entry. For all the reasons discussed above, over time small teams and creative individuals will increasingly be able to make Hollywood-quality content for pennies on the dollar- leading to what I've been calling “infinite content.” And while Hollywood is currently reeling from the disruption of distribution that Netflix triggered 15 years ago, these falling entry barriers could trigger a next wave of disruption.
+
+The silver lining for Hollywood is that these technologies can lower their costs too. So, if you're running a big studio, how can you capitalize? You're managing a large business, with a lot of people used to doing things a certain way. You are also competing for creative talent with other studios and generally don't have the
+
+https://archive.ph/WE4AQ
+
+19/22
+
+# Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
+
+4/23/25, 6:56 PM
+
+bargaining power to tell them how to do their job, especially the most sought-after A-listers. ("Yes, Chris Nolan, we love your latest project, but we will be requiring some fundamental changes in your creative process...")
+
+Adopting these new technologies will be a large challenge technologically, but it will be an even bigger change management challenge. Getting people to change is really hard. I know. That's why it will be so much easier for small independent teams, starting with a clean piece of paper, to adopt these tools much faster.
+
+For an established studio, there are two possible paths:
+
+*   Choose a non-core process to test. The most politically viable processes will be those that are already done by third-parties. For instance, you might shift localization services to AI-enabled providers in some markets or you could bring more VFX work in house with the mandate to use AI tools (and lower costs).
+*   Create a skunkworks. In this case, you would establish a separate studio to start from scratch to test the relative cost, quality and speed of "AI-first" content production.
+
+Neither of these incremental approaches are likely to move the needle a ton in the near-term, but at least they will start to build up AI "muscle memory" in the organization.
+
+## Head-Spinning, I Know
+
+All of this is moving at an dizzying pace. Even if you spend a lot of time trying to stay on top of these developments, as I do, it's hard to keep up. If you work in the industry, it may be enthralling. It may also be overwhelming and scary.
+
+For good or ill, technology marches on. Forearmed is forewarned.
+
+### Subscribe to The Mediator
+
+By Doug Shapiro
+
+The Mediator is (mostly) about the long term structural changes in the media industry and the business, cultural, and societal implications of those shifts. I write it to get closer to the frontier.
+
+By subscribing, I agree to Substack's Terms of Use, and acknowledge its Information Collection Notice and Privacy Policy..
+
+[Image of four people's profile pictures with the text "4 Likes" next to them.]
+
+[Image of a heart icon] 4 [Image of a comment icon] 1 [Image of a refresh icon]
+
+Previous
+
+Discussion about this post
+
+https://archive.ph/WE4AQ
+
+Share
+
+Next →
+
+20/22

inbox/archive/general/shapiro-cant-just-make-hits.md

@@ -0,0 +1,813 @@
+---
+source_type: "article"
+title: "You Cant Just Make the Hits"
+author: "Doug Shapiro"
+url: "https://dougshapiro.substack.com/p/you-cant-just-make-the-hits"
+date_published: "2023-04-01"
+date_archived: "2025-04-23"
+archived_by: "clay"
+domain: "entertainment"
+status: processed
+claims_extracted:
+  - "cost-plus deals shifted economic risk from talent to streamers while misaligning creative incentives"
+  - "the TV industry needs diversified small bets like venture capital not concentrated large bets because power law returns dominate"
+---
+# You Can't Just Make the Hits - by Doug Shapiro
+
+archive.today Saved from https://dougshapiro.substack.com/p/you-cant-just-make-the-hits
+search
+23 Apr 2025 17:52:16 UTC
+no other snapshots from this url
+Webpage capture
+All snapshots from host dougshapiro.substack.com
+Webpage
+Screenshot
+
+## You Can't Just Make the Hits
+
+Why the TV Business Needs to Tackle Rising Risk
+
+DOUG SHAPIRO
+APR 17, 2023
+
+[Note that this essay was originally published on Medium]
+
+share
+download.zip
+report bug or abuse
+Share
+
+The image shows a black and white abstract rendering of a professional cinema camera exploding into many small cubes. The background is a gradient of dark to light gray. The camera is positioned on the left side of the image, with the explosion emanating from it.
+
+Midjourney, prompt: "professional cinema camera exploding, black and white, clean
+background, abstract style-ar 16:9"
+
+The value of any business, or any financial instrument for that matter, is a function of
+two things: growth and risk. It has a direct relationship with the former and an
+indirect relationship with the latter.
+
+It's widely understood that in the past year growth expectations have declined in the
+TV business. What isn't as well understood is that risk is also rising. In this essay, I
+explain why TV has become riskier, why that's putting increasing pressure on returns
+in TV and what the big media companies can do about it.
+
+https://archive.ph/J88sw
+
+1/15
+
+## You Can't Just Make the Hits - by Doug Shapiro
+
+Tl;dr:
+
+*   TV and film production has always been a hit-driven business. But the model is
+    riskier than ever for three compounding reasons: spending per project has gone
+    up (duh); risk has shifted to content buyers from sellers; and the variance of
+    returns is climbing because more value is being concentrated in fewer hits.
+*   The first driver of increased risk needs little elaboration. Intuitively and
+    empirically, production cost per TV series and film has climbed in recent years.
+*   Second, risk has shifted to content buyers (streamers and networks) from sellers
+    (talent and studios) because of business practices pioneered by Netflix and
+    adopted industry-wide. These include cost-plus deal structures, massive upfront
+    overall deals for top talent and straight-to-series orders.
+*   Lastly, more value is concentrating in fewer hits for a variety of reasons: the
+    dwindling middle and lengthening tail of popularity means that the biggest hits
+    are relatively bigger than the average; hits are more global than ever; every hit is a
+    potential franchise; and, perhaps most important in a D2C environment, hits have
+    an outsized effect on subscriber acquisition (which I show with new data from
+    Parrot Analytics).
+*   The big media companies need to lower risk. The response so far-shifting
+    resources to franchises-won't solve the problem owing to franchise
+    commoditization (not “fatigue”) and the rising bargaining power of top talent.
+*   The short term solution is to revert back to historical deal structures that
+    appropriately share risk and reward with talent and independent studios. The long
+    term, and much tougher, solution is a fundamental rethinking of the risk profile of
+    video content creation.
+
+Thanks for reading The Mediator! Subscribe for
+free to receive new posts and support my work.
+
+## Growth Expectations in TV Have Fallen
+
+I won't belabor this point. It has become increasingly clear over the past year that
+streaming won't likely compensate for declining profits in traditional pay TV.
+Consumers apparently don't have an appetite for as many monthly SVOD
+subscriptions as once hoped; churn is much higher than many expected (with a
+significant proportion of subscribers regularly disconnecting and reconnecting
+depending on the content available); and content spend remains very high owing to
+both the competitive dynamic and the need to satisfy newly empowered consumers'
+insatiable demand for new content. To cap it off, the pressure on the traditional pay
+TV business also continues unabated, with the pace of subscriber losses picking up in
+recent quarters.
+
+I've written about these dynamics in several prior posts, including One Clear Casualty
+of the Streaming Wars: Profit (10/2020), Is Streaming a Good Business? (08/2022) and
+Media's Shift from Growth to Optimization (10/2022).
+
+2/15
+
+## You Can't Just Make the Hits - by Doug Shapiro
+
+Perhaps the best way to make the point is a recent chart from SVB MoffettNathanson
+showing free cash flow (FCF) for the major public media companies (Figure 1). Note
+both the stark decline from peak levels (Disney achieved peak FCF of $9.9 billion in
+F2018, not shown on the chart) and the expectation that, other than Netflix, none will
+re-achieve historical levels of FCF by 2025.
+
+Figure 1. Historical and Expected FCF for Media Conglomerates
+
+The image is a bar graph titled "Free Cash Flow by Company". The graph shows the free cash flow in billions of dollars for several media companies (DIS, WBD, NFLX, FOXA, PARA, AMCX) for the years FY19, FY22, and FY25E. The graph indicates a decline in free cash flow for most companies from FY19 to FY22, with projections for FY25E showing some recovery but not reaching FY19 levels for most.
+
+Note: Disney FCF was ~$9.9 billion in F2018. Disney on September fiscal year, Fox on June
+fiscal year. Source: SVB MoffettNathanson.
+
+The idea that free cash flow growth expectations have fallen is widely understood.
+What's less well understood is that risk has also increased.
+
+## Risk Driver #1: Higher Cost per Project
+
+I won't belabor this point either. (Don't worry, there's plenty of belaboring below.) It
+tracks intuitively that spending per project in TV (and, for that matter, movies) has
+climbed in recent years. The data also back that up.
+
+Here's a chart I showed in another recent post, Forget Peak TV, Here Comes Infinite
+TV (01/23).
+
+Ten years ago, production costs for the average hour-long cable drama were about
+$3-4 million. Today it is common to see dramas exceed $15 million per episode
+(Figure 2).
+
+Figure 2. Many TV Series Now Exceed $15 million Per Episode in Production Costs
+
+3/15
+
+## You Can't Just Make the Hits - by Doug Shapiro
+
+The image shows a bar graph titled "Highest Budget TV series per episode of all time: as of 2022". The graph shows the reported production budget in US$ millions for various TV series, including "The Rings of Power", "Stranger Things S4", "Hawkeye", "Falcon + Winter soldier", "Wandavision", "House of the Dragon", "Game of Thrones S8", "The Pacific", and "The Sandman". The budgets range from $15 million to $58 million per episode. The network or streaming service for each series is also indicated.
+
+Highest Budget TV series per episode of all time: as of 2022
+
+TV series name
+Reported production budget (US$ millions)
+Network:
+
+The Rings of Power 58 prime video
+Stranger Things S4 30 NETFLIX
+Hawkeye 25 Disney+
+Falcon + Winter soldier 25 Disney+
+Wandavision 25 Disney+
+House of the Dragon 20 HBOmax
+Game of Thrones S8 15 HBO
+The Pacific 20 HBOmax
+The Sandman 15 NETFLIX
+
+Source: Sta
+
+Here's
+an film
+n't
+t doubled.
+adjusted f
+Figure 3. T
+20 Years
+budget ha
+some grea
+
+The image shows two line graphs. The first graph is titled "Median production budgets of live-action fiction feature films". The x-axis represents the release year, ranging from 2000 to 2021. The y-axis represents the reported production budget in millions of dollars. The graph shows the median production budgets fluctuating over the years, with a general upward trend. The second graph is titled "Median production budgets of live-action fiction feature films, by budget range". It contains two line graphs, one for "$50m - $100m" and another for "Over $100m". The x-axis represents the release year, ranging from 2000 to 2021. The y-axis represents the reported production budget in millions of dollars. Both graphs show the median production budgets fluctuating over the years, with a general upward trend.
+
+Median production budgets of live-action fiction feature films
+$45
+$40
+$35
+$30
+$25
+$20
+$15
+$10
+StephenFollows.com
+$5
+
+Median production budgets of live-action fiction feature films, by budget range
+$50m - $100m
+Over $100m
+$90
+$80
+$70
+$60
+$50
+$40
+$100
+$30
+$20
+$50
+$10
+StephenFollows.com
+S-
+S-
+2000
+2001
+2002
+2003
+2004
+2005
+2006
+2007
+2008
+2009
+Release year
+2010
+2011
+2012
+2013
+$150
+$200
+2014
+2015
+2016
+2017
+2018
+2019
+2020
+2021
+
+Includes all live-action fictional feature films were released in North America on home entertainment by a distributor who typically
+represented theatrically distributed films outside of the pandemic, and for which a budget figure is available.
+Budgets in non-USD currencies were converted to USD at the rate in their principal production year. Figures not inflation adjusted.
+
+Source: Stephen Follows.
+
+## Risk Driver #2: Risk Has Shifted to Buyers
+
+There has been a structural shift of risk from talent and studios to networks and
+streamers over the past decade too. This is due to several changes in industry practices
+pioneered by Netflix that have been adopted industry-wide in recent years.
+
+Historically, when producing TV, studios (and, indirectly, talent) would bear relatively
+high degrees of risk and retain substantial upside. (Note that sometimes studios are
+independent third parties and sometimes they are owned within the same corporate
+entity as the network/streaming service. For our purposes, I am making the
+simplifying assumption that affiliated studios operate at arms length from their
+
+4/15
+
+## You Can't Just Make the Hits - by Doug Shapiro
+
+affiliated networks/streaming services and will gloss over the distinction and just use
+the word "studios.") Studios would license their shows to broadcast (and to a lesser
+degree, cable) networks at a deficit, meaning that the license fees wouldn't cover
+production costs. But studios retained backend rights, so they profited from any home
+entertainment, international licensing or syndication revenue after the initial run.
+(And, depending on the contractual relationship between the studios and the show
+runners/writers/actors, that upside was shared with talent.) That's how series like
+Seinfeld, Friends, The Simpsons or The Big Bang Theory became billion-dollar properties
+for studios and talent.
+
+When Netflix started offering original programming in 2011, it decided to eliminate
+the backend. It wanted to build its originals library to reduce reliance on licensed
+content and didn't want to license those originals to third parties. It also had global
+ambitions. As a result, it sought to retain rights to its originals for very long periods
+(generally ten years or more after the series ends), in all territories. To secure those
+rights, Netflix need a new template to compensate studios and talent. It established
+several practices, all of which shift risk to networks and streamers:
+
+*   Cost-plus structures. The most fundamental shift in deal structures was toward
+    "cost-plus deals.” Rather than license shows at a deficit, streamers agreed to pay a
+    premium over cost ("cost-plus”) of generally around 20%. Under this structure, the
+    streamers are paying a premium for all shows, whether they succeed or not. The
+    flip side is that the streamer also owns the rights when a show hits, not the studio.
+    In practice, however, this hasn't been a great tradeoff. Because they are generally
+    not licensing these shows off platform, there are no more syndication/home
+    entertainment/international windfalls; they have capped the upside. In addition,
+    generally these deals have clauses that increase talent compensation and budgets
+    (and, therefore, the absolute dollar value of the premium, which is a percentage of
+    the budget) if the series extends past a certain number of seasons. Even if this isn't
+    contractual, the talent has substantial bargaining leverage when negotiating the
+    outer seasons of a hit. A good example is Stranger Things. The first season
+    reportedly cost $6 million per episode and season four reportedly rose to $30
+    million per episode. Some of the increase was higher production values and much
+    longer run times, but it also included significantly higher compensation for the
+    stars. According to Puck, for instance, Winona Ryder will make $9.5 million for
+    season five, up from $1 million in season one.
+*   Lucrative overall deals. In an overall deal, a studio secures all of a
+    writer/producer's output for a set period of time (usually two-three years, but
+    sometimes as long as five). It pays a guaranteed fee, which is then recouped to the
+    extent the writer/producer is successful over that period. The highest profile
+    recent overall deals include Ryan Murphy ($300 million from Netflix), Shonda
+    Rhimes (reportedly worth between $300–400 million from Netflix), Tyler Perry
+    ($150 million annually plus an equity stake in BET+ from Paramount), Greg
+    Berlanti ($400 million from WarnerBros. Discovery) and JJ Abrams ($250 million
+    from WarnerBros. Discovery). While these are all as close as you get to household
+    names among showrunners, in recent years it has also become common for many
+    less well-known writers and producers to get overall deals. These deals are all
+    structured differently and the “headline” parenthetical numbers above all mean
+    something different. In some cases (Ryan Murphy), these headline numbers are
+
+5/15
+
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+guaranteed and relatively fixed, in others (Shonda Rhimes), they are structured with lower guarantees and higher incentive payments and the totals are just rough estimates. As a generality though, they include large guaranteed payments even if projects fail and therefore represent a significant risk for streamers.
+
+*   Straight-to-series orders. Prior to Netflix's entrance into original programming, common practice in show development involved ordering a pilot episode for somewhere between ~$3–10 million for a scripted hour of TV (although some pilots have run much more than that). Network executives decided whether to greenlight a season (or, often, first half of a season) based on the quality of the pilot and, sometimes, reaction of focus groups. Far less common was the "straight-to-series” order, when a network committed to an entire season, or even several seasons, sight unseen. (An exception that proved the rule was when Disney committed to a whopping 44 episodes of Steven Spielberg's Amazing Stories in 1985. But that's Steven Spielberg.) Netflix changed that in 2011 when it ordered two full seasons to win bidding for House of Cards. Since then, straight-to-season orders have become standard practice. This shift has materially changed the risk associated with ordering a new scripted show: rather than spend $5–10 million on a pilot, now it is necessary to spend $80-100 million or more on a full season.
+
+Rather than spend $5–10 million on a pilot, now it's necessary to spend $80–100 million or more on a full season.
+
+# A Brief(ish) Digression: In TV, Content is King Again
+
+The late Sumner Redstone was fond of saying "content is king." It's pithy and memorable but not categorically true. While content is arguably the most important component of the overall entertainment experience, it is only one component. Think of it this way: “Content is king” is true in the same sense that “food is king" in the restaurant business. (Service, cleanliness, ambience, location, ease of parking, etc., can all be important factors.)
+
+Non-content elements of an entertainment experience include the UI, including ease of search and quality of recommendations; fidelity (stream quality and resolution of a TV show, graphic quality in a game, bit rate of a song); breadth of supported form factors; whether or not it is interrupted by ads; and social elements, among other things.
+
+In TV, the relative importance of content has changed over time. We can think about this shift in three eras:
+
+# Content is King (1980s-2008)
+
+In the pay TV era, when Redstone first coined the phrase, content was clearly critical, because it was the only real differentiator in the TV viewing experience. Most people (~90% of households) purchased a package of cable networks through their local cable or telco operator or a national satellite provider. Everyone watched TV on a...wait for it...television, accessed all their video content through the same (usually crappy) Comcast/DirecTV/Verizon electronic program guide (EPG) and sat through 16-18
+
+[https://archive.ph/J88sw](https://archive.ph/J88sw)
+
+6/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+minutes per hour of ads. In that environment, the only differentiator in the experience of consuming TV was the program itself.
+
+# Content is (Temporarily) Dethroned (2008–2019)
+
+In the early streaming era, when most consumers supplemented their pay TV subscription with one or more SVOD services, the relative importance of content started to decline owing to the rise of new differentiators in the TV experience. These included ad-free vs. ad-supported; all on-demand vs. a mix of on-demand and broadcast; how many episodes or seasons were available on demand; a choice of new form factors; easy search, navigation and discovery (including personalized recommendations); and other advanced features (like playback markers that enabled users to start a show on one device and pick up on another, parental controls, etc.).
+
+Anytime someone came home, turned on Netflix first and then decided what to watch second, he was essentially signaling that other elements of the TV viewing experience had become more important than the content itself. When I was at Turner, we had all kinds of survey data showing that people were opting to only watch ad-free shows or would check to see whether multiple seasons were stacked before starting a new series -both indications of the declining relative importance of the content itself.
+
+# Content Returns From Exile (2019-present)
+
+Now we're in the third era, when the relative value of content has shifted back. Netflix still has a better UI than most other streamers, but its relative competitive advantage has diminished. All streaming content (on Max, Disney+, Peacock, etc.) is now available on demand, with multiple stacked seasons and, if you're willing to pay for it, ad-free. Since the overall TV viewing experience is sufficiently similar between different streaming services, the actual programming is once again the key differentiating factor.
+
+Now that other elements of the streaming experience are sufficiently similar, content is again the key determinant of quality.
+
+# Risk Driver #3: More Value is Concentrated in Fewer Hits
+
+So, while content in general has become more important and valuable, a growing proportion of that value is concentrated in fewer hits. In the language of finance, the variance of returns is increasing, and therefore risk. There are several reasons.
+
+# Fatter Head, Longer Tail
+
+This was the topic of my last essay, Power Laws in Culture. The main point was that, even in a world of near-infinite content, entertainment popularity distributions persistently, and in some cases increasingly, approximate power laws: a few massive hits and a very, very (very) long tail. As I described in that piece, this is an inherent feature of networks.
+
+[https://archive.ph/J88sw](https://archive.ph/J88sw)
+
+7/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+The hits in the head are becoming relatively bigger compared to the average show or movie.
+
+As I also described (and showed empirically), with significant (or growing) consumption in the head and an ever longer tail, the middle is getting hollowed out. So, even if they are not absolutely bigger (higher absolute viewers, constant dollar box office, etc.) the hits in the head are becoming relatively larger compared to the average show or movie.
+
+This can be seen in Figure 4, which shows the distribution of global "demand" for top Netflix series in 2018, 2020 and 2022, from Parrot Analytics. Parrot's demand metric incorporates a variety of inputs (social, fan and critic ratings, piracy, wikis, blogs, etc.) to gauge the popularity of each series and movie on each streaming service. The top chart shows the distribution for the top 250 Netflix series and the bottom zooms in on just the top 50. As shown, over time the distribution of demand is becoming even more skewed to the top hits (note how steeply the blue line drops off from the head of the curve).
+
+Figure 4. For Netflix, the Distribution of Demand for Series is Becoming More Skewed to the Top Hits
+
+The image shows two line graphs illustrating the distribution of total global demand among top Netflix series. The first graph displays the distribution among the top 250 series, while the second graph zooms in on the top 50 series. Each graph contains three lines representing the years 2018, 2020, and 2022. The x-axis represents the rank of the series, and the y-axis represents the percentage of total global demand. The graphs show that the distribution of demand is becoming increasingly skewed towards the top hits over time, as indicated by the steeper drop-off in the blue line (2022) compared to the other lines.
+
+DISTRIBUTION OF TOTAL GLOBAL DEMAND AMONG TOP 250 SERIES
+ON NETFLIX
+2018-2020-2022
+
+4.  0%
+5.  5%
+6.  0%
+7.  5%
+8.  0%
+9.  5%
+10. 0%
+11. 5%
+12. 0%
+
+DISTRIBUTION OF TOTAL GLOBAL DEMAND AMONG TOP 50 SERIES ON
+NETFLIX
+2018-2020-2022
+
+133
+39
+69
+87
+205
+
+4.  0%
+5.  5%
+6.  0%
+7.  5%
+8.  0%
+9.  5%
+10. 0%
+11. 5%
+12. 0%
+
+1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
+
+[https://archive.ph/J88sw](https://archive.ph/J88sw)
+
+Source: Parrot Analytics, Author analysis.
+
+# Globalization
+
+It has long been true that domestic (U.S.) hits have been popular internationally, in part because the size of the U.S. entertainment market justified higher investment and
+
+8/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+consequently better production values than anywhere else. In recent years, however, the reverse has also been true: there has been growing domestic demand for international hits. The result is that the biggest hits, both domestically and foreign-produced, increasingly have broad global appeal.
+
+Figure 5 shows demand data from Parrot for Netflix originals in 2022, both in the U.S. and globally. As shown, of the top 40 most-demanded series both in the U.S. and around the world, 29 were on both lists. In addition, the most-demanded shows in the U.S. included many that debuted internationally, some of which are non-English language, such as Peaky Blinders, Squid Games, Dark, Narcos, Komi Can't Communicate, La Casa De Papel and The Last Kingdom.
+
+Figure 5. There was High Degree of Overlap Among the Most-Demanded Netflix Original Series Last Year Domestically and Globally
+
+The image is a table comparing the most-demanded Netflix original series in the United States and globally in 2022, according to Parrot Analytics. The table lists the top 40 series in each category, with overlapping titles highlighted. The key indicates that titles with no overlap are not highlighted. The table shows a significant degree of overlap between the most-demanded series in the U.S. and globally, suggesting that popular Netflix originals tend to have broad international appeal.
+
+Domestic
+Global
+1 Stranger Things
+Stranger Things
+2 Cobra Kai
+Peaky Blinders
+3 The Witcher
+The Witcher
+4 Peaky Blinders
+5 Ozark
+La Casa De Papel (Money Heist)
+Lucifer
+Bridgerton
+Ozark
+Cobra Kai
+6 Lucifer
+7 Bridgerton
+8 Marvel's Daredevil
+9 Arcane
+10 The Umbrella Academy
+11 You
+12 The Crown
+13 BoJack Horseman
+14 Ask The StoryBots
+15 Snowpiercer (2020)
+16 Squid Game
+17 Black Mirror
+18 Dark
+19 Orange Is The New Black
+20 Love Death + Robots
+21 Komi Can't Communicate
+22 Love
+23 La Casa De Papel (Money Heist)
+24 Castlevania
+25 Lost In Space
+26 Big Mouth
+27 The Dragon Prince
+28 Disenchantment
+29 Narcos
+30 The Last Kingdom
+Arcane
+Squid Game
+Marvel's Daredevil
+The Crown
+Black Mirror
+Love Death + Robots
+The Queen's Gambit
+The Umbrella Academy
+Dark
+Sex Education
+Narcos
+All of Us Are Dead
+The Last Kingdom
+Komi Can't Communicate
+House Of Cards
+Alice in Borderland
+Emily In Paris
+Snowpiercer (2020)
+Formula 1: Drive To Survive
+Shadow And Bone
+You
+Lost In Space
+13 Reasons Why
+31 Shadow And Bone
+32 One Day At A Time
+33 The Queen's Gambit
+34 Longmire
+35 Storybots Super Songs
+36 Emily In Paris
+37 Shopkins
+38 Marvel's The Punisher
+BoJack Horseman
+Castlevania
+Mindhunter
+Love
+Sweet Home
+Orange Is The New Black
+Kingdom
+39 She-Ra And The Princesses Of Power Space Force
+40 Grace And Frankie
+Sacred Games
+Key
+No Overlap
+
+Source: Parrot Analytics.
+
+# Hits are Extensible
+
+As I discuss below, in an bid to attract viewers who are overwhelmed by choice, studios have been allocating more resources toward developing "franchises” that revolve around familiar IP.
+
+Clearly, IP with rich mythology-Game of Thrones, Lord of the Rings, the MCU, Harry Potter, etc. offers almost limitless opportunities for prequels, sequels, reboots and auxiliary story lines. But in recent years, the definition of franchise has broadened; anything that's considered a hit is now a potential franchise. As recent examples, Yellowstone has spawned three spinoffs, 1883, 1923 and 6666; and Amazon and Michael B. Jordan are reportedly exploring a “Creed-verse” that would include multiple film and TV projects.
+
+[https://archive.ph/J88sw](https://archive.ph/J88sw)
+
+Every hit is a latent franchise.
+
+9/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+Plus, successful franchises can also be extended into other experiences and products, like gaming, theatrical, live events and merchandise. Netflix recently announced an animated spinoff of Stranger Things and a Stranger Things play and VR game are both expected later this year.
+
+# Hits Disproportionately Drive Subs
+
+Hits have always been important. In traditional ad-supported pay TV, for instance, a hit show draws more viewers- which directly increases advertising revenue-and creates a brand halo that draws viewers to other programming on a network and helps attract talent.
+
+But hits are even more important in a direct-to-consumer environment because they have a disproportionate impact on attracting subscribers. Over the last 12–18 months, it has become evident that one of the TV industry's biggest surprises and biggest problems is high streaming churn. (See To Everything, Churn, Churn, Churn.) Attracting and retaining subscribers are streamers' top priorities and biggest challenges.
+
+It's pretty intuitive that the biggest hits are the biggest drivers of subscriber additions. For empirical evidence, let's look at more Parrot data. In addition to tracking demand for each title, Parrot also tracks the programming that viewers watch both before and after they view each title. As a result, Parrot can estimate to what degree each series or movie attracts new subscribers (i.e., the preceding title viewed is on a different streaming service) or helps retain subscribers (i.e., the preceding title viewed is on the same streaming service).
+
+Figure 6 shows the proportion of both demand and gross adds represented by the top 10 titles on Apple TV+, Amazon Prime Video, Disney+, HBO Max, Hulu, Paramount+, Peacock and Netflix in 1Q23. As shown, these titles represented a large portion of demand (10-50%) and a much larger proportion of gross additions (50–80%).
+
+Figure 6. The Vast Majority of Gross Adds are Tied to the Top 10 Titles
+
+The image is a bar graph comparing the share of gross adds and share of demand derived from the top 10 exclusive titles on various streaming platforms in the U.S. during the first quarter of 2023. The x-axis lists the streaming platforms: Amazon Prime Video, Apple TV+, Disney+, HBO Max, Hulu, Netflix, Paramount+, and Peacock. The y-axis represents the percentage, ranging from 0% to 100%. For each platform, there are two bars: one representing the share of gross adds and the other representing the share of demand. The graph shows that the top 10 exclusive titles generally account for a larger proportion of gross adds than of demand across all platforms, indicating that these titles are more effective at attracting new subscribers than reflecting overall viewer interest.
+
+PROPORTION OF DEMAND AND GROSS ADDS
+DERIVED FROM TOP 10 EXCLUSIVE TITLES IN
+1Q23, U.S.
+Share of Gross Adds
+Share of Demand
+
+100%
+90%
+80%
+70%
+60%
+50%
+40%
+30%
+20%
+10%
+0%
+
+Amazon Prime Apple TV+ Disney+
+Video
+HBO Max
+Hulu
+Netflix
+Paramount+ Peacock
+
+Source: Parrot Analytics.
+
+# The TV Business Needs to Reduce Risk
+
+[https://archive.ph/J88sw](https://archive.ph/J88sw)
+
+10/15
+
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+As mentioned at the beginning, the value of any business or financial instrument is a
+function of growth and risk (of cash flows). There is a direct relationship for the former
+and an indirect relationship for the latter. When risk goes up, value goes down. For
+liquid public securities, like stocks or public debt, prices immediately fall when
+perceived risk rises. Anyone who has ever done a discounted cash flow analysis knows
+that the net present value of a company is highly sensitive to the debt and equity risk
+premia embedded in the weighted average cost of capital. In other words, risk matters.
+A lot.
+
+Mitigating risk is just as important as reinvigorating growth.
+
+The big media companies have recently taken several steps to boost growth, like price
+increases (from Netflix and Disney), new ad-supported tiers (also Netflix and Disney),
+some signs of moderation in the pace of content spend, a crackdown on password
+sharing (Netflix), combination of subscale services to bolster subscriber growth (the
+combination of Paramount+ with Showtime and HBO Max with Discovery+). But
+rising risk is also putting increasing pressure on returns. Mitigating risk is just as
+urgent as reinvigorating growth.
+
+A Shift to Franchises Won't Work
+
+Big media's initial attempts at risk mitigation have included allocating more
+development spend to franchises, as mentioned before. As documented in this great
+article, a growing proportion of hit movies and TV shows (as well as other media) are
+derivative content (prequels, sequels, reboots, etc.). Ampere Analysis also found that
+64% of SVOD originals in 1H22 were based on pre-existing IP. But allocating more
+resources to franchises probably won't meaningfully change the risk profile for a
+couple of reasons:
+
+Franchise commoditization. Many observers bemoan the growing prevalence of
+franchises and the concept of “franchise fatigue" periodically rears its head, especially
+whenever there is a string of unsuccessful franchise extensions (such as recently
+occurred at Disney, with disappointing results for Andor, The Mandalorian season three
+and Ant-Man and the Wasp: Quantumania). Whether franchise fatigue is a valid concern
+is an open question. For every Ant-Man disappointment there is a hit like John Wick 4
+around the corner. The implication is that people want quality entertainment,
+franchise or not. The bigger issue is not fatigue, however, it is commoditization. The
+premise behind increased allocation of development towards franchises is that, in a
+crowded marketplace, familiar IP attracts viewers and moviegoers. The problem is
+that everyone is pursuing the same strategy. It may not be a race to the bottom, but it
+is a race to the familiar. When everything is a franchise, franchises no longer stand out.
+
+Franchise fatigue isn't the issue; franchise commoditization is the issue.
+
+High degree of talent bargaining leverage. The other challenge with franchises is that
+talent often has substantial bargaining power when negotiating franchise extensions.
+
+https://archive.ph/J88sw
+
+## 11/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+The lead actors for Batman and James Bond may be (somewhat) fungible, since these
+franchises have swapped actors many times. Other are non-negotiable, like Tom
+Cruise in Mission Impossible 7 or Top Gun: Maverick, Daniel Craig in Knives Out, Vin
+Diesel in Fast X, the cast of Stranger Things or Taylor Sheridan (showrunner of
+Yellowstone and its spinoffs). These stars (and their agents) are well aware that their
+involvement is critical or sometimes required for a sequel/prequel/reboot to proceed
+and can extract huge upfront payments and profit participations as a result.
+
+Given the talent costs, "low-risk” franchises aren't really low risk.
+
+A Short-Term Approach: Share Risk with Talent
+
+So, if franchises aren't the solution, what is? The most obvious short run solution is a
+reversion back to historical deal structures that transfer more risk (and potential
+reward) to talent and studios. This includes a reduction in overall talent deals (or at
+least tying them more closely to success) and straight-to-series orders. There are signs
+this is happening. In fact, Netflix recently reportedly ordered its first pilot ever.
+
+The biggest change would be a shift away from cost-plus deals to better align
+producers' and distributors' interests. Netflix has taken an initial step in this direction
+and is reportedly trying to move premiums to flat rate fees, rather than percentage
+premiums. A full step would entail lower premiums, and possibly even deficits, in
+exchange for re-instituting backend participation.
+
+The challenge here, of course, is that it's difficult to provide backend incentives when
+most streamers have been reluctant to license to third parties and there still is no
+backend. One option is to create a “synthetic” backend formula (based on viewership
+and perhaps other metrics) to calculate and share backend value with talent. Given the
+pressure on the business and the growing evidence that the full value of content is not
+being realized when constrained to only one window (i.e., SVOD), it is also
+increasingly likely that streamers ultimately re-embrace licensing (see Media's Shift
+from Growth to Optimization).
+
+Netflix hasn't done this yet, but there is growing willingness from the traditional
+media companies. WarnerBros. Discovery has been vocal about its openness to
+licensing and recently struck a deal to license content to Roku and Tubi. At a recent
+investor conference Disney CEO Bob Iger also said that the company was re-
+evaluating making content for third parties. As a possible early indication of this, last
+month Netflix announced that Arrested Development, which is owned by Disney and
+was originally slated to leave the service, will stay on after all.
+
+A Long-Term Approach: Fundamentally Rethink “Portfolio
+Construction" in TV
+
+The industry could conceivably reverse some of the disadvantageous deal structures
+that it has adopted in recent years (risk driver #2). But what can it do about structurally
+higher variance of returns (risk driver #3)?
+
+Throughout this essay, I've touched on a few financial topics, like risk and variance.
+Let's turn to another one: diversification. When professional investors construct a
+
+https://archive.ph/J88sw
+
+## 12/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+portfolio, they don't just care about the expected returns, they care about the expected
+returns per unit of risk, or risk adjusted returns. (The intuition here is that you'd much
+rather invest in a portfolio with 20% expected upside and 10% potential downside than
+20% expected upside and 50% potential downside.) Modern Portfolio Theory (MPT)
+(which is not so modern, since it was formulated in 1952) dictates that the way to
+reduce the risk of a portfolio is by adding low correlation investments.
+
+Under MPT, the higher the average variance of the investments in a portfolio, the
+more low correlation investments you need to produce a given level of risk. This is
+why, for instance, a private equity fund (which tends to buy relatively stable, cash
+flowing businesses) might construct a portfolio with 10-15 investments, while a
+venture capital fund (which invests in much higher risk, earlier stage companies, about
+half of which usually fail) invests in 20-40 companies, or more.
+
+The TV business needs to think more VC, less PE.
+
+To bring it back to TV, to lower risk, the TV industry needs to think more VC, less PE:
+it needs a more diversified approach. The implication is that the studio of the future
+should look much different than the studio of today. Here's a rough sketch of what that
+might mean:
+
+*   More shots on goal at much lower cost, facilitated by new technologies. In light
+    of the increasingly skewed return distributions of content, studios need to take
+    many more shots on goal, at much lower cost. Fortunately, as I discussed a few
+    months ago (Forget Peak TV, Here Comes Infinite TV), this will become
+    increasingly feasible over the next several years as AI-enhanced and assisted
+    production tools evolve and proliferate. Within the relatively near term, it should
+    be possible for smaller creative teams to make very high quality content with
+    significantly smaller budgets and shorter time frames. History dictates that the
+    performance curve will improve very quickly from there. Over the longer term (5+
+    years), will it be possible to make high quality content for an order of magnitude
+    less, or even more? When you consider that the technological gating factors are
+    the sophistication of algorithms, size of datasets and compute power, the answer
+    is probably yes. For some vivid examples of what these technologies can already
+    do, check out this running Twitter thread:
+
+*   Social as a development tool, not a marketing tool. Today, studios view social
+    networking as a marketing tool to be leveraged once a show is deep in
+    development or in the can. In the future, however, it will make sense to seed pilots
+    onto "the network" (YouTube, TikTok, etc.) to see which ideas surface and which
+    don't-and then develop the successful concepts and discontinue those that fail to
+    attract attention.
+
+*   Better alignment between talent and streamer. Another way to enable more shots
+    on goal is a much more equitable sharing of risk and reward with talent. As
+    described above, today development is incredibly expensive and risky,
+    necessitating that the streamers (with millions of subscribers and billions of
+    dollars of revenue) shoulder most of the risk and retain most of the reward. If the
+
+https://archive.ph/J88sw
+
+## 13/15
+
+# 4/23/25, 6:56 PM
+
+You Can't Just Make the Hits - by Doug Shapiro
+
+cost of development plummeted, however, this would no longer be necessary. With
+much lower development costs, it would probably be advantageous to share rights
+(and therefore profits) much more equally with creatives to incent them to create
+the best possible product at the lowest possible cost.
+
+*   Creatives and technologists on an equal footing. In a studio today, there is a very
+    clear hierarchy. Creatives (or the development executives who nurture the
+    relationships with creatives) get the corner office and technologists lurk in the
+    basement pining away for a little sun. In the modern (or post-modern) studio,
+    creatives and technologists would have more equal status. Staying on top of fast-
+    moving technology will be almost as critical as producing the most compelling
+    content.
+
+Easy to Say, Hard to Do
+
+As with many of the things I've written recently, the main point is that the TV and
+film businesses have reached an inflection point and many of the old rules will
+(eventually) need to at least re-evaluated, if not torn up and re-written.
+
+That's easy for me to say, of course, but it will be extraordinarily hard to do. The major
+media companies are part of a large and complex creative ecosystem of talent (both the
+highly successful and those struggling to make a living), guilds, trades and agencies.
+(As just one topical example, it is worth noting that in its pending contract
+renegotiation, the Writers' Guild of America (WGA) is reportedly seeking to constrain
+studios' ability to use AI.)
+
+There are many disparate and often conflicting vested interests in Hollywood,
+sometimes with cinematically-large egos, and getting them all to march in time will be
+an enormous challenge. But progressive executives will have to try.
+
+Subscribe to The Mediator
+By Doug Shapiro
+
+The Mediator is (mostly) about the long term structural changes in the media industry and the business,
+cultural, and societal implications of those shifts. I write it to get closer to the frontier.
+
+By subscribing, I agree to Substack's Terms of Use, and acknowledge
+its Information Collection Notice and Privacy Policy.
+
+[Previous](None)
+
+Discussion about this post
+
+Comments Restacks
+
+[Share](None)
+
+[Next →](None)
+
+https://archive.ph/J88sw
+
+## 14/15

inbox/queue/.extraction-debug/shapiro-ai-use-cases-hollywood.json

@@ -0,0 +1,35 @@
+{
+  "rejected_claims": [
+    {
+      "filename": "genai-adoption-in-entertainment-gated-by-consumer-acceptance-not-technology.md",
+      "issues": [
+        "missing_attribution_extractor"
+      ]
+    },
+    {
+      "filename": "non-atl-production-costs-converge-with-compute-costs-as-ai-replaces-labor.md",
+      "issues": [
+        "missing_attribution_extractor"
+      ]
+    }
+  ],
+  "validation_stats": {
+    "total": 2,
+    "kept": 0,
+    "fixed": 5,
+    "rejected": 2,
+    "fixes_applied": [
+      "genai-adoption-in-entertainment-gated-by-consumer-acceptance-not-technology.md:set_created:2026-03-19",
+      "genai-adoption-in-entertainment-gated-by-consumer-acceptance-not-technology.md:stripped_wiki_link:AI-labor-displacement-follows-knowledge-embodiment-lag-phase",
+      "non-atl-production-costs-converge-with-compute-costs-as-ai-replaces-labor.md:set_created:2026-03-19",
+      "non-atl-production-costs-converge-with-compute-costs-as-ai-replaces-labor.md:stripped_wiki_link:AI-labor-displacement-follows-knowledge-embodiment-lag-phase",
+      "non-atl-production-costs-converge-with-compute-costs-as-ai-replaces-labor.md:stripped_wiki_link:two-phase-disruption-where-distribution-moats-fall-first-and"
+    ],
+    "rejections": [
+      "genai-adoption-in-entertainment-gated-by-consumer-acceptance-not-technology.md:missing_attribution_extractor",
+      "non-atl-production-costs-converge-with-compute-costs-as-ai-replaces-labor.md:missing_attribution_extractor"
+    ]
+  },
+  "model": "anthropic/claude-sonnet-4.5",
+  "date": "2026-03-19"
+}

inbox/queue/.extraction-debug/shapiro-cant-just-make-hits.json

@@ -0,0 +1,44 @@
+{
+  "rejected_claims": [
+    {
+      "filename": "cost-plus-deals-shift-risk-from-talent-to-streamers-while-misaligning-creative-incentives.md",
+      "issues": [
+        "missing_attribution_extractor"
+      ]
+    },
+    {
+      "filename": "tv-industry-needs-diversified-small-bets-like-venture-capital-not-concentrated-large-bets-because-power-law-returns-dominate.md",
+      "issues": [
+        "missing_attribution_extractor"
+      ]
+    },
+    {
+      "filename": "franchise-commoditization-not-fatigue-is-the-strategic-problem-when-everyone-pursues-familiar-ip.md",
+      "issues": [
+        "missing_attribution_extractor"
+      ]
+    }
+  ],
+  "validation_stats": {
+    "total": 3,
+    "kept": 0,
+    "fixed": 7,
+    "rejected": 3,
+    "fixes_applied": [
+      "cost-plus-deals-shift-risk-from-talent-to-streamers-while-misaligning-creative-incentives.md:set_created:2026-03-19",
+      "cost-plus-deals-shift-risk-from-talent-to-streamers-while-misaligning-creative-incentives.md:stripped_wiki_link:giving-away-the-commoditized-layer-to-capture-value-on-the-s",
+      "tv-industry-needs-diversified-small-bets-like-venture-capital-not-concentrated-large-bets-because-power-law-returns-dominate.md:set_created:2026-03-19",
+      "tv-industry-needs-diversified-small-bets-like-venture-capital-not-concentrated-large-bets-because-power-law-returns-dominate.md:stripped_wiki_link:industries-are-need-satisfaction-systems-and-the-attractor-s",
+      "tv-industry-needs-diversified-small-bets-like-venture-capital-not-concentrated-large-bets-because-power-law-returns-dominate.md:stripped_wiki_link:AI-optimization-of-industry-subsystems-induces-demand-for-mo",
+      "franchise-commoditization-not-fatigue-is-the-strategic-problem-when-everyone-pursues-familiar-ip.md:set_created:2026-03-19",
+      "franchise-commoditization-not-fatigue-is-the-strategic-problem-when-everyone-pursues-familiar-ip.md:stripped_wiki_link:disruptors-redefine-quality-rather-than-competing-on-the-inc"
+    ],
+    "rejections": [
+      "cost-plus-deals-shift-risk-from-talent-to-streamers-while-misaligning-creative-incentives.md:missing_attribution_extractor",
+      "tv-industry-needs-diversified-small-bets-like-venture-capital-not-concentrated-large-bets-because-power-law-returns-dominate.md:missing_attribution_extractor",
+      "franchise-commoditization-not-fatigue-is-the-strategic-problem-when-everyone-pursues-familiar-ip.md:missing_attribution_extractor"
+    ]
+  },
+  "model": "anthropic/claude-sonnet-4.5",
+  "date": "2026-03-19"
+}

inbox/queue/shapiro-ai-use-cases-hollywood.md

@@ -7,10 +7,14 @@ date_published: "2023-09-01"
 date_archived: "2025-04-23"
 archived_by: "clay"
 domain: "entertainment"
-status: unprocessed
+status: null-result
 claims_extracted:
   - "GenAI adoption in entertainment will be gated by consumer acceptance not technology capability"
   - "non-ATL production costs will converge with the cost of compute as AI replaces labor across the production chain"
+processed_by: leo
+processed_date: 2026-03-19
+extraction_model: "anthropic/claude-sonnet-4.5"
+extraction_notes: "LLM returned 2 claims, 2 rejected by validator"
 ---
 # 4/23/25, 6:56 PM Al Use Cases in Hollywood - by Doug Shapiro - The Mediator
 
@@ -548,3 +552,18 @@ Share
 Next →
 
 20/22
+
+
+## Key Facts
+- Median blockbuster film budget is approximately $200 million (2023)
+- Avatar 2: The Way of Water production costs exceeded $400 million
+- Barbie production budget was $145 million
+- Game of Thrones final season cost $15 million per episode
+- Lord of Rings series cost over $25 million per episode
+- Avengers: Infinity War involved approximately 4,500 people
+- Avatar: Way of Water had 98% of shots requiring VFX
+- Average VFX spend on big-budget films is ~$50 million, reaching $100 million for effects-heavy films
+- Film production budgets typically break down as: 15-20% above-the-line, 50% below-the-line, 25-30% post-production, remainder other
+- Approximately 2/3 of film production costs are labor
+- Runway Gen-2 increased video generation length from 4 seconds to 18 seconds
+- Disney used GenAI to create the title sequence for Secret Invasion

inbox/queue/shapiro-cant-just-make-hits.md

@@ -7,10 +7,14 @@ date_published: "2023-04-01"
 date_archived: "2025-04-23"
 archived_by: "clay"
 domain: "entertainment"
-status: unprocessed
+status: null-result
 claims_extracted:
   - "cost-plus deals shifted economic risk from talent to streamers while misaligning creative incentives"
   - "the TV industry needs diversified small bets like venture capital not concentrated large bets because power law returns dominate"
+processed_by: leo
+processed_date: 2026-03-19
+extraction_model: "anthropic/claude-sonnet-4.5"
+extraction_notes: "LLM returned 3 claims, 3 rejected by validator"
 ---
 # You Can't Just Make the Hits - by Doug Shapiro
 
@@ -811,3 +815,14 @@ Comments Restacks
 https://archive.ph/J88sw
 
 ## 14/15
+
+
+## Key Facts
+- Average hour-long cable drama production costs rose from $3-4M per episode ten years ago to commonly exceeding $15M today
+- The Rings of Power cost $58M per episode, Stranger Things S4 $30M per episode
+- Median film production budgets have roughly doubled over 20 years (not inflation-adjusted)
+- Netflix ordered its first pilot ever in 2023, reversing straight-to-series practice
+- Disney CEO Bob Iger said company is re-evaluating making content for third parties
+- WarnerBros Discovery struck deals to license content to Roku and Tubi
+- Writers Guild of America is reportedly seeking to constrain studios' ability to use AI in pending contract renegotiation
+- SVB MoffettNathanson projects only Netflix will re-achieve historical FCF levels by 2025 among major media companies