Generation Modes

AutoVio supports two modes for how a video is conceived and styled: style_transfer and content_remix.

style_transfer

Goal: Reuse the visual style and structure of an existing video with new content.

You upload a reference video in the Init step.
In Analyze, vision AI extracts: scenes, tone, color palette, tempo, camera movements, text overlays.
That analysis is passed into Scenario so the LLM generates scenes that match the reference’s style and structure.
You can save the analysis as the project style guide so all works in the project can use that look and feel.

Use when: You have a reference ad or template and want to create a new version with different product or messaging but the same style and pacing.

content_remix

Goal: Create a new video from text only, without a reference video.

No reference video is required. You describe the video via product name, description, target audience, duration, scene count.
Analyze is skipped (or optional if you later add a reference).
Scenario uses only user intent (and optional project style guide and knowledge) to generate scene-by-scene prompts.
Image and video generation still use the project/work style guide and instructions if set.

Use when: You’re starting from an idea or brief and don’t have a reference clip, or you want the AI to invent structure and style from your description.

Comparison

Aspect	style_transfer	content_remix
Reference video	Required (upload in Init)	Not required
Analyze step	Run to get structure & style	Skipped (or optional)
Scenario input	Analysis + intent	Intent (+ optional style guide)
Typical use	“Like this video, but for my product”	“Make a video from this description”

Mode in the API

When creating or updating a work, set mode to "style_transfer" or "content_remix". The same field is used in scenario and analyze requests so the backend and prompts can adapt (e.g. analyzer prompt focuses on style for style_transfer and on content/story for content_remix).

See Projects & Works and Pipeline Overview.