Guide
ComfyUI Longer Videos with Sustained Motion
How to create 10+, 15, or 20-second videos where motion flows continuously rather than repeating every 5 seconds.
What You'll Learn
- Why video models default to repeated/looping motion
- Frame chaining technique for continuous motion
- Multi-clip workflows for scene transitions
- Prompt strategies for longer videos
🔄 Why Motion Repeats
Most video models (Wan 2.1, LTX-Video) are trained on 5-second clips. When you generate a longer video:
- The model repeats its "learned" 5-second pattern
- Action restarts at ~5 second intervals
- You'll see visible "cuts" or loops at 5s, 10s, 15s marks
⚠️ This is a fundamental limitation. Wan 2.1 is trained on 5s videos. Going beyond that requires special techniques.
🛠️ Methods for Sustained Motion
| Method | Length | Difficulty | Best For |
|---|
| Frame Chaining | 10-20s | Medium | Continuous motion |
| Multi-Clip | 15-30s | Hard | Scene transitions |
| LTX-Video Native | 8-12s | Easy | Quick longer videos |
| Infinite/Loop Workflow | Unlimited | Advanced | Very long sequences |
🔗 Method 1: Frame Chaining (Recommended)
Generate the first 5 seconds, then use the last frame as the input for the next segment.
Steps:
- 1Generate your first clip (5 seconds, e.g., 16fps = 80 frames)
- 2Extract the last frame from the output
- 3Use that frame as the input image for I2V generation
- 4Write a prompt that continues the action:
continuing from previous scene, [action]...
- 5Repeat until you have the desired length
- 6Concatenate all clips in a video editor
Key Tips:
- Use the same seed for consistency
- Match frame rate across all clips
- Add
seamless transition to your prompt
- Overlap by 1-2 frames for smoother joins
Best Practice: Keep prompts simple and action-focused. "A person walking forward" is easier to continue than complex scene changes.
🎬 Method 2: Wan 2.1 Multi-Clip
Use the Wan 2.1 MultiClip workflow to generate longer videos (up to 20 seconds) in a single pass.
Where to Find:
- ComfyUI menu:
Workflow → Browse Templates → Video → Wan 2.1 MultiClip
- Or search for "Wan 2.1 I2V MultiClip" in ComfyUI Manager
How It Works:
- Load a starting image
- Set total frames (e.g., 16fps × 15s = 240 frames)
- The workflow splits into multiple "clips" internally
- Each clip uses the previous clip's last frame as context
⚠️ VRAM Intensive: MultiClip requires ~16GB VRAM for longer videos. May not work on 12GB cards.
✍️ Prompt Strategy for Sustained Motion
Instead of:
A person running across a field, jumping over a fence, and rolling
Do this (frame chaining):
- Clip 1:
A person running across a field
- Clip 2:
Continuing, the person approaches a fence and jumps
- Clip 3:
After jumping, the person continues running and rolls
Prompt Templates:
Continuing from previous frame, [action continues]Seamless continuation, [new action]Following the previous scene, [character] then [action]
Negative Prompts (Important):
worst quality, inconsistent motion, blurry, jittery, frozen, repeating, loopcut, fade, transition glitch
📋 Quick Reference: Frame Count
| Duration | 16 fps | 24 fps | 30 fps |
|---|
| 5 seconds | 80 frames | 120 frames | 150 frames |
| 10 seconds | 160 frames | 240 frames | 300 frames |
| 15 seconds | 240 frames | 360 frames | 450 frames |
| 20 seconds | 320 frames | 480 frames | 600 frames |
⚠️ Quality Warning: Even with these techniques, videos longer than 10-15 seconds often show quality degradation, motion artifacts, or consistency issues. Consider breaking longer narratives into multiple shorter clips.