Step 1 - Init

Step 1: Init

The Init step is where you configure the work and, for style transfer, upload the reference video.

Fields

Field	Description
Source	Reference Video — upload a video to analyze; Start from Scratch — generate from text only.
Mode	style_transfer — replicate reference video style; content_remix — reinterpret content. Only shown when a reference video is selected.
Product name	Subject of the video (e.g. product or topic name).
Product description	Optional short description or video idea.
Target audience	Who the video is for.
Resolution	Output resolution for image/video generation. See presets below.
Language	(Optional) Output language (e.g. en, tr).
Video duration	(Optional) Desired total duration in seconds.
Scene count	(Optional) Number of scenes to generate.
Reference video	(Source: Reference Video only) Upload the video to analyze.
Project assets	(Optional) Select assets to use in video generation.
Asset usage mode	How to use selected assets: reference or direct.

Resolution

The resolution selector offers three presets that control the output dimensions for all image and video generation in this work:

Preset	Width × Height	Icon	Description
Portrait 9:16	1080 × 1920	📱	Default. Vertical format for mobile / social.
Landscape 16:9	1920 × 1080	🖥️	Horizontal widescreen format.
Square 1:1	1080 × 1080	⬛	Square format.

Choosing a resolution affects:

• DALL-E 3 — selects the matching size parameter (1024x1792, 1792x1024, or 1024x1024).
• Runway — adds a ratio field ("768:1280" or "1280:768"). Square is omitted (Runway does not support 1:1).
• Gemini Veo — adds an aspectRatio field ("9:16", "16:9", or "1:1").
• Gemini Image — no effect (provider has no resolution parameter).
• Editor export settings — exportSettings.width and exportSettings.height are initialized from the chosen resolution when opening the editor for the first time.

Reference video

• For style_transfer, upload a reference video in this step. It is stored under the work and used in Step 2: Analyze.
• For content_remix, you can leave this empty; the Analyze step will be skipped unless you add a reference later.

Project assets

You can select images from your project assets to use in video generation. When assets are selected, choose an asset usage mode:

Mode	Description
Reference	AI analyzes your assets and generates similar-looking images. Best for style consistency. Assets should have descriptions (analyze them first in Project Settings).
Direct	Skip AI image generation entirely. Use your actual asset images and only generate video (animation). Best for product photos, screenshots. Scene count auto-adjusts to match asset count.

This is useful for:

• Product demo videos using actual product photos (direct mode)
• Brand-consistent content using logo/style reference (reference mode)
• Screenshot-based tutorials (direct mode)

What gets saved

• All form fields are saved on the work (mode, productName, productDescription, selectedAssetIds, assetUsageMode, resolution, etc.).
• The reference video file is stored via POST /api/projects/:id/works/:id/media/reference.

Next step

Go to Step 2: Analyze to run vision AI on the reference video (or skip to Step 3: Scenario if you have no reference).

See Pipeline Overview and Step 2: Analyze.