Fewer warped hands and faces: what to change in the prompt
The prompt-side fixes for the classic generation failures (melting hands, drifting faces, extra limbs), ordered by how often each one works.
Apply in this order until the artefact is gone:
1. Remove the second action. One beat per clip.
2. Hide the hands: 'hands in pockets', 'holding {object} with both hands', 'hands out of frame'.
3. Change the shot size one step wider (close-up → medium) or one step tighter (medium → close-up); the in-between sizes fail most.
4. Make the camera static.
5. Add at the end: 'anatomically correct, five fingers, single person, no extra limbs, no duplicate faces'.
6. Shorten to 4-5 seconds.
7. Switch to image-to-video from a clean still of the same shot. Paste into the video model as one paragraph; remove the labels once you are happy with the content.
Tips
- Step 1 fixes more clips than steps 2-6 together.
- A negative prompt is the last resort, not the first: on most current models it changes little, and on some it degrades the shot.
| works with | Veo: Google text-to-video / image-to-video prompts. · Kling: Kling text/image-to-video prompts. · Sora: OpenAI video generation prompts. · Runway: Runway Gen prompts and camera controls. |
| cost | Free. The video or image model you paste it into has its own pricing. |
| status | Unverified: written from the technique, not yet run end to end by us. |
| license | CC BY 4.0 for the text here; skills are MIT. |
| updated | 2026-11-01 |
| for agents | fix-warped-hands-and-faces.md |
Related
Image-to-video: generate the first frame, then animate it
Make a still with an image model first, fix the composition there, and only then ask the video model to animate it. Cheaper iterations and the look you chose, not the one the video model guessed.
Camera movement vocabulary that video models understand
The 14 camera moves that generation models respond to reliably, with the phrasing that works and the phrasing that produces a wobbling mess.
Anime style: 2D cel look, limited animation, clean line weight
A filled 7-part prompt for a hand-drawn anime shot: flat cel shading, animation on twos, a held painted background, and the three things the style breaks in current video models.
Architectural walkthrough: an interior on a steadicam in natural light
A filled 7-part prompt for a real-estate or design walkthrough: one slow steadicam path through a room, windows as the only light, straight verticals, and the turns that make geometry drift.
Generate a 10-clip b-roll pack for a talking-head video
One agent prompt that reads your transcript and writes ten generation prompts for cutaways that show what is being said, each tagged with the sentence it covers.
Establishing shot: open a video with a place
A prompt for the wide opening shot that tells the viewer where they are, with the three things that make a generated landscape look filmed rather than rendered.