A Patchwright-Style AI Film Workflow for Image Planning and Video Tests
A Patchwright-style workflow starts before video. It starts with images that define what the world is allowed to be.
Write the world bible
Define premise, object, material list, room rule, character rule, and camera rule. Keep it short enough to reuse.
Make design plates
Create stills for the room, main object, hands, material close-up, and final frame. These plates become the continuity source for video tests.
Write shot contracts
For each plate, decide what is locked and what can move. A good contract is short: keep brass tool, blue thread, worn canvas; only motion is tool opening and thread tightening.
Run motion tests
Test object motion, hand contact, and room continuity. Pause outputs and inspect material stability before judging style.
Compare models fairly
Use the same plates and contracts across models. The useful metric is not charm. It is usable seconds, stable objects, and low repair time.
Build the world bible as a constraint system
A Patchwright-style workflow starts with constraints, not adjectives. Write the premise, the main object, the material list, the room rule, the character rule, and the camera rule. Then add one negative rule: what must never appear. That last line matters. It prevents the model from filling gaps with franchise costumes, random logos, famous faces, or decorative machinery that breaks the world.
A good one-page bible might say: the room repairs memories through fabric; the main tool is brass with glass needles; the repeated visual motif is blue thread-light; every surface has been repaired once; the camera notices hands before faces; no readable brands, no celebrity likeness, no unrelated steampunk clutter.
Plate-to-shot pipeline
- Room plate: establish the workshop geometry and light direction.
- Object plate: lock the hero tool, material, scale, and silhouette.
- Hand plate: define how the performer touches the tool.
- Material plate: show the repaired fabric or patched surface.
- Final plate: define the emotional destination of the sequence.
Only after those plates exist should video generation begin. Otherwise the model has to invent art direction and motion at the same time, which is where many AI shorts fall apart.
Shot contracts for motion tests
A shot contract should be shorter than a prompt pack. It names the locked elements and the allowed movement. For example: keep the brass tool, glass needles, blue thread-light, worn canvas table, and warm side light; only motion is the tool opening and the thread tightening; no extra fingers, no text, no logo, no new props.
Run three tests: object motion, hand contact, and room continuity. Pause each output. If the tool changes shape, go back to the object plate. If the hand floats, simplify contact. If the room changes style, strengthen the world bible.
Revision comes after stability
Do not test edits until the base shot is stable. Once it is stable, try one revision at a time: change light temperature, slow the hand motion, remove a background object, or tighten the camera move. A model that can preserve the object through revision is much more useful than a model that only creates a beautiful first pass.
GPT Image angle: the still stage should carry more weight
For an image-first workflow, the still stage is not a mood board afterthought. It is where ownership, character identity, material rules, and shot endpoints are decided. Strong image plates make later video tests fairer because each model receives the same visual evidence instead of inventing a new world from text alone.
FAQ: when the workflow is ready
How many shots should I test first? Start with three. More shots create more failure surfaces before you know the model's limits.
When should I abandon a prompt? After the same failure appears three times with small variations. At that point, the prompt is not under-specified; the shot may be too complex or the reference too weak.
What should I save for the next project? Save the scene bible, winning reference frames, rejected outputs with failure notes, and the final edit rules. A good AI video workflow leaves behind reusable production memory.
What is the biggest mistake? Asking the model to solve story, design, camera, action, identity, lighting, and editing in one generation. Split the job. The workflow becomes slower at the beginning and faster by the end.
The full workflow beyond the prompt
A Patchwright-style workflow needs five layers before the final edit. The first layer is the world bible. The second is the design plate set. The third is the shot contract. The fourth is motion testing. The fifth is revision testing. If any layer is skipped, the model has to invent missing structure, and the short starts to feel like a collection of unrelated generated clips.
The world bible defines what exists and what never appears. The design plates show the room, the tool, the hands, the repaired surface, and the final emotional image. The shot contract defines what stays locked and what moves. Motion testing asks whether the model can animate the contract. Revision testing asks whether the model can change one thing without destroying everything else.
Shot contract examples
Object motion contract: keep the brass tool, glass needle tips, blue thread-light, worn canvas table, and warm side light. Only motion is the tool opening and the thread tightening. No text, no logo, no new props, no extra fingers.
Hand contact contract: keep the tool scale, hand position, sleeve material, and table texture. Only motion is the thumb pressing a latch and the thread catching. Cut before the fingers become unreadable.
Room continuity contract: keep the same wall repairs, same light direction, same workbench, and same blue thread motif. Only motion is a slow camera slide from the tool to the repaired wall.
Revision tests
Once the base motion passes, test revision one variable at a time. Change the light from warm to cooler. Remove one distracting background object. Slow the hand by twenty percent. Tighten the camera move. If the object survives these edits, the workflow is becoming useful. If every edit breaks the object, the model may be good at generation but weak at controlled filmmaking.
How GPT Image 2 should use it
For GPT Image 2, the strongest contribution is upstream: character sheets, object plates, final frames, and visual contracts that make later video tests fair. Keep the same plates and contracts across tests. The winning workflow is the one that preserves the designed world through generation, revision, and final edit.
Operating the workflow like a production test
A workflow article should be detailed enough that a reader can run the test without guessing the missing steps. The creator should know what to prepare before generation, what to write in the prompt, what to inspect after each output, and when to stop. The stop rule is especially important. Many AI video workflows waste time because the creator keeps regenerating after the same failure has already repeated. A practical rule is simple: if the same failure appears three times after small prompt changes, change the reference, simplify the shot, or remove one moving element.
The workflow should also describe what to save. Save the winning prompt, but also save the failed prompt, the reference frame, the failure reason, the number of attempts, and the seconds kept. Those notes are not paperwork. They are the memory that makes the next project cheaper. A creator who only saves the final clip has to rediscover the entire workflow later.
For GPT Image 2, this means the still-image stage should carry enough structure that video testing becomes less random. That project lens should appear as a testing method, not as a forced sales pitch. The reader should understand how to translate the same scene bible, shot contract, and edit rules into this product's model or workflow context. The goal is to make the article useful even for someone who is still comparing tools.
Final production checklist
Before calling the workflow done, check five things. The character remains recognizable in motion. The scene has one clear camera intention. The main object or costume detail survives at least one cut. The final clip can be published without relying on a famous likeness or protected world. The retry budget is written down. If those five checks pass, the workflow is no longer just a prompt experiment. It is a repeatable production pattern.
The best AI video workflows are not the loudest ones. They are the ones that keep creative intent visible while reducing ambiguity. They tell the model less at once, but they tell it with better structure.
Asset handoff notes
One more practical layer is the handoff between design and generation. The person preparing references should label the room plate, object plate, hand plate, material plate, and final frame with the exact rule each one is meant to protect. If the room plate protects light direction, say so. If the object plate protects brass shape and glass needle tips, say so. If the final frame protects emotional destination, say so. This makes the later video test easier to diagnose because every failure can be traced to a specific asset or instruction.
For GPT Image 2, the same handoff habit keeps the article from becoming generic model commentary. The model is not judged in a vacuum. It is judged against a prepared set of visual obligations. If the output breaks the object, the object plate needs work or the model cannot hold it. If the output preserves the object but loses pacing, the edit contract needs work. That distinction is exactly what a production workflow should teach.
Takeaway
Image planning is not a warm-up. For AI film, it is the structure that gives video generation something coherent to animate.
