Film treatments / pitch decks will become the next interface for AI Filmmaking

I often encounter discussions on co-creating films with AI in natural language, akin to collaborating with an artist, art director, or producer, and narrating a story orally. From my experience in pre-production for numerous films (through Ambushed.tv, a treatment and pitch deck agency I cofounded 4 years ago), I believe this perspective is somewhat limited and requires a new angle: exploring our current collective creation of complex visual narratives and how it will evolve with AI.

An image is worth a thousand words

Henrik Ibsen famously said, 'A thousand words leave not the same deep impression as does a single deed' that then became, 'An image is worth a thousand words.' I'd add that an image combined with copy is probably worth a hundred thousand words. Consider how children's books efficiently teach new concepts, now rivaled by videos - now, let me tell you about why film treatments otherwise known as pitches, might become a great interface to communicate with generative AI tools.

Planning a film treatment or pitch deck requires narrative visual mapping

Here is an example of collaborative "mapping" of a film treatment or pitch deck before layout that allowed us to visualize our film's rhythm and the overall visual impact of the different scenes:

Collaborative creativity is challenging

Creating top-tier films today involves multiple stakeholders from diverse cultural and technical backgrounds. The key is effectively pitching the evolving idea to keep everyone aligned and contribute value to a coherent product. Traditional film treatments or pitch decks, once simple copies with a few images, became ultra-complex world-building efforts from specialist teams as words alone are often insufficient for a film. It's like the Tower of Babel: uniting diverse minds to create an ambitious film. And now, how will we, as a collective, collaborate with AI to craft visual experiences as complex as human and machine 'minds'?

Creating means using a variety of tools

For films, a one-size-fits-all solution seems impractical. We need the freedom to play with different tools and combine them for originality. This applies to AI tools too, as emphasized by Emad Mostaque of Stability AI: in his vision, each tool will have its specialists, further fragmenting the creative process and possibly affecting the final output's coherence. From our experience, treatments are already used by the whole industry as technical documents before shootings to align the whole crew and artists (subjective visions) with the goal concept of the film.

Storyboards are visual-only boards that inspire the most recent treatment's layout

In the treatment and pitch deck industry, we saw a growing trend towards horizontal, one-line documents that bridge the gap between traditional "page-by-page" formats or vertical website-like scrolling through.

Below is an example we did, building both the scene's blockings on Blender and displaying a "map" of the actors in our scene, then composing the pictures with AI to create the world, scene by scene, and reach a mix of efficient storyboarding and pitching for this music video:

Optimizing AI results

AI, seemingly simple, responds better when guided precisely. For instance, when I introduced ChatGPT 3 to writers, basic prompts yielded basic results. However, detailed requests improved the output significantly. Precise interaction with AI is crucial. It requires prompting with words, positive and negative, but also with visual cues that can be ultra-precise (3D shadows or shapes on which to apply textures and faces) or more global and connected to the overall universe (Lora models of a certain art style, product, character or lightning effects…). On top of this, the effect of Creative Chaos comes from this creative ping-pong with the machine (sometimes amazing surprises and sometimes, very poor results).

Let's take an example with one interesting interface (Comfy.UI with Stable Diffusion), to generate a very short chaotic hallucination but with a subject that we control with both text and images:

Visual cue:

Text prompt (positive and negative):

Creating this "guided Chaos" generation for trial and demonstration purposes only (note how the machine is "trying" at different stages what it thinks of our combination of prompts and visual cues):

This all looks very familiar to people who create film treatments! Now, imagine being able to communicate with much more detail with more complex and agile technological stacks, scene by scene, in a coherent, cinematic way… This is what we feel is coming!

Film treatments or pitch decks as the next AI co-creation interface

A film treatment (in the advertising world), or pitch deck (in the fiction industry), detailing a film's story and style, is vital for communicating a film's vision. In an AI-collaborative future, we believe that treatments will change and become key for discussing visual creations collaboratively before "pitching" the film to different AI generators that will be computing what we mean cinematically. While AI tools will rapidly evolve, our approach to collaborative creation will remain constant and human-based, which calls for the most efficient ways of pitching complex visuals and cinematic experiences. In short, in our (frankly biased) opinion, they're now the most efficient interface for co-creating a universe, story, and film.

At synest.ai, we can help you build a fully operational and creative pipeline, in-house, to prepare for the next Generative AI improvements.