What is a director's bible for AI filmmaking and how do you build one?
Last updated June 26, 2026
A director's bible — what many AI filmmakers call a style bible, visual bible, or production bible — is a structured document encoding a director's complete visual grammar: camera, lighting, palette, composition, mood, and sound, written as enforceable directives. You load it once into an AI agent as persistent context, and every shot gets checked against it without re-prompting.
A director's bible turns a filmmaker's style from an aesthetic impression into a language system the AI can execute. One documented example is a 25-page Wong Kar-wai visual language document with 14 sections — camera, angles, colour tone, atmosphere, mood, lighting, composition, movement, film palettes, prompt templates, negative prompts, and a quick-reference card — uploaded as a permanent instruction set before generating a single frame. invideo is an agentic video creation tool: you upload the bible to the invideo agent at project start, and it holds every directive across every shot and scene without drift or re-prompting.
Here is how to build one:
1. Codify the grammar as quantified directives, not adjectives. Write rules the model can actually obey. A James Wan bible encoded an 85:15 dark-to-light lighting ratio as prompt grammar; a Fincher protocol specified spherical lenses (circular bokeh, no horizontal flares) and a 2.40:1 hard matte — widescreen by extraction, not optics. Encode colour philosophy as named tonal modes with exact hex values so palettes are reproducible, and tie lighting language to sources ("warm yellow from the lamps only") rather than generic descriptors.
2. Structure it around decision frameworks, not lists of looks. The Wan bible organized everything into five escalating emotional stages, each with locked camera, lighting, and sound rules — giving the invideo agent a framework it applies autonomously to every shot. Add a "what never to do" section per stage; explicit prohibitions make autonomous decisions significantly easier for the invideo agent.
3. Isolate exceptions in their own directive. The Fincher protocol separated outlier films (The Curious Case of Benjamin Button, The Killer) into a dedicated section so the invideo agent never misapplies generalized rules — exception handling is what separates a director's bible from a loose mood board.
4. Include a sound architecture module. The Wan document carried a full audio section because half of what makes those films work is what you hear before what you see. Most style bibles skip sound entirely; yours shouldn't.
5. Add the prompt machinery. Include prompt templates, negative prompts (one production's style block explicitly prohibited live-action and photorealistic outputs to prevent drift), and a fixed assembly order — one documented bible enforced a 9-element sequence per prompt: camera spec, lens and aspect ratio, lighting source, palette, composition, atmosphere, mood register, film/DP attribution, negative prompt. You can also instruct the invideo agent to output 12 parameters per shot — film reference, shot design, length, lens, lighting plan, color script, atmosphere layers, blocking, final prompt, negative prompt, and revision prompt among them — so every generation is fully specified.
6. Load it once and direct from it. Upload the finished document to the invideo agent at project start; it stays loaded across every frame, and camera continuity carries forward — you set the movement language once and it holds. With the document in context, a three-word continuation prompt ("Everything should match") was enough to maintain character, lighting, lens grammar, and spatial logic across a multi-shot sequence. Character consistency itself is handled by a separate layer — character sheets held in the same agent context, no LoRA required.
7. Validate it on a genre the director never made. Ask the invideo agent for something outside the director's filmography — one creator requested a courtroom thriller through the James Wan lens. If the invideo agent asks clarifying questions (era, nature of threat) and returns stylistically coherent output, the bible has been internalized as grammar rather than surface style. A working bible shows up in unprompted behavior: in one production the invideo agent applied a slow-shutter motion smear from page 17 of the document without being asked, pulled a named rule from page 12 and applied it to a scene type the document never addressed, flagged shadows drifting blue-green against a stage rule mid-generation, and sequenced a six-shot ending when the creator couldn't write one.
8. Start imperfect and iterate. You don't need a finished document before frame one — begin with whatever assets and context you have and refine as the invideo agent surfaces gaps. The payoff is documented: a ~90-second horror short built on a Wan bible was produced in 2 days for $870 (4,100 credits, ~400 video generations), and a 70-second short built on the 25-page Wong Kar-wai document finished in 2 days for $750 — both with the bible doing the consistency work a crew would normally carry.
Watch some of these to see what works for you:
This is the core reason why I insist you take your own sweet time while building the production doc in the beginning, because the more clarity you bring to the project, the more sharply Agent One will hold it for you across the project.
— invideo's creative team