AI lip sync video for full dialogue scenes

Two ways lip sync gets made

There are two broad approaches. One renders a video first and then re-animates the mouth to match an audio track afterward. The other has the video model produce the picture and the speech together, so the mouth is generated with the sound. Newer video models with native audio take the second approach.

ShowMaker uses the video models' native audio for lip-synced dialogue. The character's voice is fixed on their record, so the same voice speaks every line, and you choose per shot which model renders it.

That choice matters. Seedance, Wan, Veo, Kling and MiniMax are all available, and they differ in how they handle speech, motion and cost. You don't have to commit a whole scene to one of them. A close-up and the wide around it can go to different models.

Why lip sync drifts

Most bad lip sync comes from a mismatch between the line and the shot. A line runs four seconds and the shot is three, so the end is clipped or the performance is rushed. Or a line is attributed to the wrong speaker and the wrong mouth moves. Or captions are timed to the script instead of the take.

These are all things you can catch before rendering. ShowMaker checks every shot and flags a line that is too long for the shot's length. The free animatic plays the voices against the stills at their real durations, so you hear the scene's rhythm before spending on video.

Speaker attribution is worth a second look in any scene with more than two people. When a line is attached to the wrong character, the wrong mouth moves and the wrong voice speaks. Re-rendering won't fix that. Correct it in the Script and the shots that use it follow.

• If a line is longer than the shot, split it or extend the shot.
• If a line has the wrong speaker, fix the attribution in the Script.
• If captions are off, caption the final edit instead of the script.

Framing that helps the mouth read

Lip sync is most visible in close-ups and medium close-ups, and least visible in wides. Put your money where the audience can see the mouth. A wide where the speaker is small can often use a cheaper model, or play the line over a reaction shot of the listener.

Multishot panels cover several shots of a scene in one take, cut on the dialogue. That keeps one continuous performance across the cuts. You can trim the take or break it back into shots afterward.

Give the performance room. A line that starts the instant a shot begins, or ends right on the cut, tends to feel clipped. A beat of silence before or after, or a reaction from the listener, lets the speech breathe and makes small timing errors much harder to spot.

Cost and attempts

The cost of lip-synced dialogue depends more on how many times you re-render each shot than on the price of one clip. Approve the script, the voices and the animatic first to keep attempts low. The free plan gives you 200 credits, no card, for narrated films from stills. Paid plans start at $19 a month for 900 credits.

Spend where the mouth is visible. Speaking close-ups and medium close-ups are worth a premium model with native audio. A wide where the speaker is small, or a line played over the listener's reaction, usually isn't. Mixing models within one scene is normal, and it keeps dialogue scenes affordable.

As one example of what a whole dialogue-heavy episode costs: a full 67-second vertical micro drama with a consistent cast and 8 multishot panels came in under $7 in credits.