🎤 How to make an AI lipsync video
A still portrait plus a voice clip is now enough to ship a talking-head ad, a tutorial intro or a fake-interview hook. Here is the exact workflow.
Get the right inputs
You need ONE clear front-facing portrait (eyes open, mouth closed, no heavy occlusion) and ONE clean audio file under 30 seconds. WAV or MP3 work best. Background music should be light or absent.
Pick a lipsync model
Lumineer ships several — pick the one matched to your need: photoreal portraits, stylised avatars or fast drafts. Each model has a sweet spot, look at the example outputs before picking.
Generate the voice first (optional)
No real recording? Generate the voice with an AI TTS model: pick a voice, paste your script, tune emotion and pacing. Then feed the resulting audio into the lipsync model.
Polish the output
If the head feels static, add a tiny camera move in post (slow zoom-in). If the eyes feel dead, regenerate with a slightly different reference photo where the subject was mid-blink.
Ready to put this into practice?
Open the studio and apply what you just learned in under a minute.
Try AI lipsync