What does the shaking head prompt have to get right?
The shaking head prompt has to do six jobs at once: keep your subject's identity from one photo, switch from squint to wide eyes in a single frame, shake side to side instead of nodding, repeat three loops on the beat, cope with animals whose eyes work differently, and leave the sound for later. On AI Shaking Head it is already written and tested. Make the video from one photo.
Unofficial: not affiliated with TikTok, CapCut, ByteDance or the original creator. This page does not print a prompt on purpose, because a pasted prompt is the least reliable part of the process.
What are the six jobs?
- Identity from one photo. The face, fur or hair, markings and outfit must stay yours through a fast head turn. This is where most AI video drifts.
- The instant switch. The original goes from eyes shut to eyes wide in about 0.3 seconds, one frame of snap. Models like to blend slowly.
- Side to side, not a nod. "Shaking head" means the "no" shake. Without clear direction a model often nods instead.
- Three loops on the beat. About 3.6 seconds per loop, three times in 10 seconds, in time with the original sound. See the beat sheet.
- Animals without obvious eyelids. A squint reads on a cat. It needs different handling for other subjects.
- Sound added after. The model renders silent video. The trend's sound is laid on top once the clip is done, so the prompt must not try to create audio.
Why do copied prompts fail?
- Wrong face. The prompt describes a cat on a bed. Your photo is a dog in a garden, and the model tries to satisfy both.
- Slow, soft motion. A long wordy prompt gives a gentle sway, not a snap. The comedy dies.
- Missing inputs. A prompt is one ingredient. The photo, the framing and the format matter just as much.
- One lucky render. A prompt that worked once for its author may fail four times for you, and each attempt on a paid model costs money.
If you expected a one-line fix, the honest answer is that there is not one. The tutorial shows the shorter route.
What is built in here?
Each of the 7 templates carries its own tested shot list and settings, tuned to its subject: Smug Cat, Dog Head Shake, Selfie Head Shake and the rest. You upload a photo and pick a format. Nothing is shown and nothing is typed. If you are curious about the app route, the CapCut template page compares it honestly, and the trend page explains where the clip came from.
