Skip to main content
Framesail AI
Docs

Styles & consistency

Analyze reference videos into a style

Style analysis reverse-engineers your references into editable text fields — art style, narrative style, director style, and text style — written as prompt-ready instructions that the downstream pipeline reads directly. The result is a topic-agnostic blueprint, not a summary: it never references the subject matter, people, or events in your sources, so the same style works for any topic you make videos about.

What it reads

Up to 12 references, in any mix: images, YouTube links, and text snippets. Each reference can be scoped to what it should teach — tag it to the art, narrative, or director fields, or leave it untagged to inform all of them. YouTube references are validated for total length when you submit.

What it produces

All four are plain, editable text — read them, tighten them, or rewrite them entirely. See Edit style fields and add custom variables for what else a style can carry.

Re-analyzing

Analysis runs as a background job, and you can re-run it anytime with new or edited references. Re-analysis overwrites only the analyzed style fields — any custom fields you added yourself survive untouched.

Related

The style analysis page shows the feature in action; if you would rather start from a curated look, see Choose style presets.