The Problem
Keeping visual style, lighting, and camera seeds consistent across a multi-scene AI video is notoriously hard. When everyone prompts their own way, scene four doesn't look like scene one, and nobody can say which seed made the good take.
| What breaks | Why |
|---|---|
| Style drifts between scenes | Each prompt is written fresh |
| Lighting doesn't match | Small wording changes shift the whole look |
| Good takes can't be reproduced | The seed wasn't saved |
| Every person prompts differently | No shared formula across the team |
How It Works
WhatThePrompt enforces parameterised prompt formulas and node recipes across the creative team. You pick a preset, fill in its parameters, and the matching ComfyUI workflow runs; the seed goes into Supabase, and the preview renders straight in the browser.
Key Decision: Formulas, Not Free Text
The core idea is that a look is a formula with blanks, not a paragraph someone retypes.
| Ad-hoc prompting | A WhatThePrompt preset |
|---|---|
| A paragraph, rewritten each time | A fixed formula with named parameters |
| Settings chosen per person | A shared ComfyUI node recipe |
| Seeds in someone's notes, if anywhere | Seed history stored in Supabase |
| Preview after a full render elsewhere | Preview in the browser |
The Result
| For the team | What it means |
|---|---|
| Consistency | Scenes in one video share a look by default |
| Reproducibility | Any good take can be traced back to its seed |
| Speed | People fill in parameters instead of writing prompts |
WhatThePrompt is an internal tool, so there's no public build or source to link to.
