Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.talkcut.com/llms.txt

Use this file to discover all available pages before exploring further.

Reference video lets you generate clips where the AI maintains the look, style, and appearance of subjects from your reference images. Provide 1-3 images as visual anchors, describe what should happen, and the AI creates a video that stays true to those references. This is how you achieve continuity across shots — reuse the same reference images, and each generated clip maintains consistent characters, environments, and visual style.

​
How reference images work

Reference images guide the AI’s visual output. The AI analyzes your references and preserves their key visual characteristics in the generated video:
  • Character appearance — face, clothing, body type, and distinctive features
  • Environment style — architecture, lighting, color palette, and atmosphere
  • Prop details — shape, color, texture, and proportions
  • Visual style — artistic approach, rendering quality, and overall aesthetic

​
Creating a reference video

1

Select reference images

Choose 1-3 images from your canvas or asset library. These define the visual identity for your video. You can use generated images, uploaded photos, or a mix of both.
2

Describe the action

Tell the AI agent what should happen in the shot — camera movement, character action, environment changes. Use filmmaking language for precision: “slow dolly in”, “rack focus to background”, “character turns to face camera”.
3

Generate

The AI creates a video clip at 720p or 1080p that maintains visual consistency with your references.

​
Best practices for reference images

  • Use clear, well-lit images — avoid noisy, blurry, or heavily compressed references
  • Show the full subject — a full-body character shot works better than a cropped headshot
  • Use a clean background — solid or simple backgrounds help the AI isolate the subject. Use Remove background if needed.
  • Match your target style — if your project is photorealistic, use photorealistic references. If it’s illustrated, use illustrated references.
  • One subject per reference — each reference image should feature a single character, prop, or scene — not a group
Generate your reference images with TalkCut first, then use those as references for video. This gives you the most consistent results because the AI understands its own outputs best.

​
Building continuity across shots

To maintain visual consistency across multiple video clips:
  1. Create a reference sheet — generate or upload clear images of each key character and environment
  2. Reuse the same references — select the same reference images each time you generate a new shot with that character or setting
  3. Use frames to organize — group reference images for each character or scene in a frame on the canvas
  4. Be consistent with style direction — use the same style descriptions across shots, or save them as a skill
By reusing consistent reference images and style directions, you achieve continuity across shots.

​
Specs

SpecValue
Aspect ratio16:9 (widescreen)
Resolution720p (default) or 1080p
AI-generated audioOptional
Duration5-8 seconds per clip
Cost840 credits per clip
Video generation is credit-intensive. Check your credit balance before generating long sequences of reference video clips.