| name | ai-avatar-video |
| description | AI avatar video guide. Talking head animation, image-to-video workflow, face animation tips. |
| allowed-tools | [] |
| disable-model-invocation | false |
AI Avatar Video Guide
Use /pocket-knife:image-to-video to animate a portrait or character image into a talking-head or expressive avatar video.
The skill uses fal.ai Wan 2.2 5B, which accepts a static image and a motion prompt, returning an MP4.
How Image-to-Video Works for Avatars
The model interprets the source image and applies motion guided by your text prompt. For avatar use:
- Subtle head movements, blinking, and lip movement are achievable with targeted prompts
- Full body movement works best when the character is clearly isolated from the background
- Realistic results require a high-quality, well-lit source image
Ideal Source Image Properties
| Property | Requirement |
|---|
| Resolution | 512×512 minimum; 1024×1024 or higher preferred |
| Subject | Clear face or full body, well-lit |
| Background | Simple or solid — busy backgrounds reduce animation quality |
| Format | JPEG or PNG hosted at a public HTTPS URL |
| Pose | Neutral, forward-facing for talking head; side views work for profiles |
The image must be a publicly accessible HTTPS URL — local file paths are not supported. Upload the image to a CDN, cloud storage bucket, or image host first.
Motion Prompt Formula for Avatars
[Facial/body action] + [Style descriptor] + [Camera behavior] + [Technical quality]
Examples:
| Effect desired | Prompt |
|---|
| Talking head, professional | "person speaking naturally, subtle head nods, lips moving, professional lighting, cinematic" |
| Animated character | "character blinking and smiling, slight head tilt, warm lighting, smooth animation" |
| Dramatic avatar | "person delivering a speech, confident gestures, dramatic lighting, cinematic close-up" |
| Subtle life animation | "gentle blinking, micro-expressions, natural breathing movement, soft studio light" |
| Fantasy avatar | "elf character turning head slowly, glowing eyes, magical ambient light, fantasy style" |
Workflow: Static Portrait to Avatar Video
Step 1 — Prepare the source image
- Use
/pocket-knife:ai-image-generation to generate a portrait with a specific character design
- Or use
/pocket-knife:background-removal to isolate a real person on a clean background
- Upload the result to a public image host (e.g., Imgur, Cloudinary, or an S3 bucket with public access)
Step 2 — Animate with image-to-video
- Run
/pocket-knife:image-to-video with the public image URL
- Use an avatar-specific motion prompt (see examples above)
- Generation takes 1–3 minutes; the skill polls automatically
Step 3 — Add voice (optional)
- Generate speech with
/pocket-knife:elevenlabs-tts using a voice that matches the avatar's character
- Layer audio over the video in any video editor (CapCut, DaVinci Resolve, Premiere)
Face Animation Tips
For realistic talking motion:
- Include
"lips moving" or "speaking" explicitly in the prompt
- Add
"natural head movement" for lifelike micro-motion
- Avoid over-specifying —
"nodding vigorously" often produces distortion
For cartoon or illustrated characters:
- Add
"smooth animation" and "cartoon style consistent" to maintain style
- Illustrated characters with exaggerated features animate more predictably than photorealistic ones
For minimal motion (looping background avatar):
- Use
"very subtle breathing motion, gentle blinking, nearly static, ambient movement only"
- This works well for profile pictures, loading screens, and background presenters
Common Issues
| Issue | Cause | Fix |
|---|
| Face distorts mid-video | Complex background distracting the model | Use background-removal before animating |
| No lip movement | Prompt too vague | Add "lips moving", "speaking" explicitly |
| Stiff, no motion | Low-contrast source image | Use sharper, higher-contrast portrait |
| Wrong motion applied | Ambiguous prompt | Be specific: "slight head turn to the right" not "move" |
| Image URL rejected | Private or expired URL | Re-upload to a public host with permanent link |
Use Cases
| Use case | Source image | Motion prompt |
|---|
| Course presenter avatar | AI-generated instructor | "speaking to camera, natural nods, professional" |
| Brand mascot animation | Illustrated character | "blinking, friendly smile, slight wave, cartoon" |
| Social media avatar | Product photo + person | "looking at camera, subtle smile, modern aesthetic" |
| Game character cutscene | Fantasy character art | "dramatic head turn, intense expression, cinematic" |
Workflow with Other Skills
/pocket-knife:ai-image-generation — generate the portrait
/pocket-knife:background-removal — isolate subject if needed
/pocket-knife:image-to-video — animate the portrait
/pocket-knife:elevenlabs-tts — generate matching voice narration
- Combine video + audio in your editor of choice