Make any photo speak

AI lip sync turns a still image and an audio clip into a video of the subject speaking or singing that audio. The model listens to the sound, works out the mouth shape each syllable needs, and animates the face to match, with the small head movements and expressions that make speech look real instead of a mouth flapping on a photo.

Mage's Lipsync app does this from one picture. Upload a photo or choose a saved character, add a voice line, a song or a narration of up to 15 seconds, and pick a video model. The subject keeps the look, framing, lighting and setting of your image while the camera holds steady on them. You can add a short scene direction, such as a new background or a mood, and Mage applies it after its own lip sync instruction.

It is built for creators who already have the audio: a voiceover for a product clip, a dubbed line in another language, a podcast quote, a verse of a song. If you still need the voice itself, make it first with Seed Audio and bring it over. If the person should only speak, Lipsync is the tool. If they should perform, with camera moves and dancing, the Music Video app is the better fit.

The audio decides most of the result. A dry recording with one voice, no music bed and no room echo gives the model clear syllables to follow. Background music and crosstalk blur the consonants that mouth shapes are built from, so trim the clip to the line you need and cut the silence from the start, since the face waits while nothing is said.

The image decides the rest. A face that fills a good part of the frame, looks roughly toward the camera and has the mouth unobstructed gives the model room to work. Hands over the mouth, hard side profiles and very small faces in a wide shot are the usual reasons a clip looks off. If you are starting from a generated portrait, make it with the mouth closed or relaxed, so the first frames do not begin mid-word.

• Trim the audio to the exact line, 15 seconds at most
• Use one clear voice without a music bed
• Choose a front-facing photo with the mouth visible
• Keep scene directions short: a place, an outfit or a mood
• Try a second model before rewriting anything

Each Lipsync run takes up to 15 seconds of audio. For a full voiceover or a longer song, split the audio at natural pauses into pieces of 15 seconds or less and sync each piece against the same photo or saved character. A saved character keeps the face identical from one segment to the next, which is what makes the cuts invisible.

When the segments are ready, line them up in Storyboard, which puts up to six saved clips in order and exports them, with their sound, as one MP4. Starting from a blank portrait instead? The AI talking photo guide covers writing the voice line as well as animating the face.

Lip sync is a practical way to dub a short clip. Record or generate the line in the new language, keep the delivery close to the original pacing, and run it against a still of a speaker you have permission to use. Because the mouth is animated from the new audio, the lips follow the translated words rather than the original ones.

For a recurring narrator or a character who speaks across a series, build the face once with the AI character generator, then add moving shots between the talking ones with AI image to video. For a sung performance with camera moves, the AI music video generator is built for exactly that.