Idle Motion Loop Animator - スタンバイ状態の動きをつける — Open Skill

Idle Motion Loop Animator - スタンバイ状態の動きをつける

Idle Motion Loop Animator - スタンバイ状態の動きをつける is a public Open Skill on Digen. Open it to see what this workflow creates.

@koara

Anime

7 0 4

https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4278658/2026-10-19/82ew8v5y.png

---

allowed-tools:
  - recognize_image
  - generate_image
  - generate_video
---

# Idle Motion Loop Animator

## When to use
- The user wants to turn a still image (uploaded, or described in text) into a short, seamlessly looping animation with subtle idle motion.
- Any art style is fine: illustration, painting, photo, 3D, etc.

## Inputs
- Image: either (a) one uploaded image, or (b) a text prompt describing the image.
- Optional: the motion the user wants in plain words; motion strength (subtle / stronger); length (5s or 10s, default 10s).

## Steps
1. Establish the reference image:
   - If an image is uploaded: call `recognize_image(image_url=<asset_id or URL>, prompt="Describe the subject, clothing/props, and background in short natural language so identity can be kept consistent.")` and note the description.
   - If only a text prompt is given: first generate a still image with `generate_image(model='t2i', prompt=<the user's description>, orientation=<match the user's intent, default 'landscape'>, name='Reference still', entity_type='reference_images')`, then use that image as the reference.
2. Plan the motion:
   - If the user gave no motion idea, suggest 2–3 motions that fit the image (e.g. hair sway, clothes ripple, prop movement, background/light flicker) and ask the user to pick one. Do not generate the video until the user has chosen.
   - Otherwise turn the idea into a motion plan of continuous gentle motion throughout, with no stop or freeze at any point, so the motion reads as a loop.
3. Generate the seamless loop video:
   - Call `generate_video(model='i2v', input_urls=[<reference image asset_id>], duration=<5 or 10, default 10>, prompt=<short description from step 1> + <motion plan> + <fixed rules below>, name='Seamless loop', entity_type='final_videos')`.
   - Use the reference image as the single start frame (no forced end frame). Forcing an identical last frame makes the model freeze to match it, so instead let the motion run freely and return to the original pose only at the very end.
   - Motion strength: if the user did not specify, default to gentle but clearly visible motion (not near-still); stronger = more pronounced but still looping.
4. Reply with the generated video and note which motion was used.

## Fixed rules (always include in the video prompt)
- Static camera; no change of angle or composition.
- Preserve the original image's style exactly (line work, coloring, texture, lighting, level of realism); no style change.
- No text, no dialogue, no new objects, no distortion of faces or hands.
- The motion must continue continuously and visibly throughout the entire duration, including the final seconds — never stop, pause, or freeze at any point, and especially do not freeze near the end. Do not hold a static pose in the middle or at the end. The subject should still be in gentle motion at the very end; make the motion periodic so it naturally returns to the original pose at the final moment for a seamless loop. The strength of the motion follows the motion strength setting.

## Notes
- If the generated video still freezes near the end, re-run with an even stronger emphasis on continuous motion, or generate a slightly longer clip and trim the frozen tail with `extract_clip`.
- Report failures honestly; never fabricate URLs.
- Politely decline requests for prohibited content.