How to make AI characters talk: the best lip sync tools and methods in 2026
A character who speaks is a character people care about. Here are the four ways AI creators make their characters talk, the tools for each, and the tricks that make dialogue scenes believable.
Four ways to make a character talk
There is no single lip sync button. Depending on the shot, creators use one of four methods: generate the dialogue together with the video, sync an existing clip to a recorded line, animate a talking portrait from a still image, or transfer a real actor's performance onto the character.
- "We leave at dawn."
- Voice track
- Character
- Lips follow every syllable
1. Dialogue generated with the video
The newest video models can generate speech and lip movement at the same time as the image. You write the line in the prompt, and the character says it. It is the fastest way to make a dialogue scene, and the results are often natural.
The trade-off is control: you cannot easily change one word or reuse the exact same voice across a whole season. Many creators use this for tests and short exchanges, and keep a separate voice track for important scenes.
- Google Flow (Veo) (Small free allowance · Google AI Pro about $20/month): Veo generates dialogue, sound effects and ambience together with the image.
- Kling AI (Small daily free allowance · from about $10/month): Kling can generate speech with the video, and also sync a clip to your own audio.
2. Lip sync on an existing clip
You already have the shot and a recorded line (for example from ElevenLabs). A lip sync tool rewrites the mouth movement so it matches the audio. This keeps one consistent voice per character for the whole series, which is what most serious creators want.
- Kling AI (Usage-based, on top of a plan): Lip sync for characters it generated or that you upload, best on clean, front-facing shots.
- Sync (Free trial · paid plans and API): A specialist known for accurate lip sync on existing footage, including real video.
- ElevenLabs (Free 10,000 credits/month (no commercial license) · Starter about $6/month): Generate the voice lines first: one consistent voice per character, in dozens of languages.
3. A talking portrait from a still
For a close-up of a character speaking to the camera, a narrator or a confession scene, you can animate a single image with an audio track. The tool generates the lip movement, the head motion and the expressions together.
- Hedra (Free plan · paid plans available): Its Character models turn a still and a voice line into an expressive talking character.
4. Transfer a real performance
For the most natural acting, film yourself saying the line with your phone, then transfer your facial expressions and gestures onto the character. You become the actor, and the AI becomes the costume.
- Runway (125 free credits · from $15/month): Act-Two transfers the expressions and gestures of a filmed performance onto a character image or video.
Which method for which shot
| Shot | Best method | Why |
|---|---|---|
| Quick exchange between two characters | Dialogue generated with the video | Fastest, natural timing |
| Main character, recurring voice | Lip sync on an existing clip | Same voice all season |
| Narrator or close-up confession | Talking portrait from a still | Expressive face, simple setup |
| Emotional key scene | Performance transfer | Real acting, subtle expressions |
Tips for believable dialogue
- Front-facing or three-quarter shots sync best. Profiles and fast motion are hard.
- One speaker per shot. Cut to the other character for the reply.
- Keep lines short. Two short shots are more convincing than one long speech.
- Use clean audio with no music under the voice, and add the music in the edit.
- Cut to reactions: a listener's face often says more than the speaker's mouth.
- Never make a real person say something without their consent.
FAQ
What is the best AI lip sync tool?
It depends on the shot. Kling and Sync are strong for syncing an existing clip to your audio, Hedra for a talking character from a single image, Veo for dialogue generated with the video, and Runway Act-Two for transferring a real performance.
Is there a free AI lip sync tool?
Several tools have free trials or plans with limits and watermarks, including Hedra, Sync and Kling's daily allowance. For a published series, paid plans remove watermarks and include commercial use.
How do I make two AI characters talk to each other?
Generate one shot per line, each with only one character speaking, then cut between them like a film dialogue scene. Use a separate, consistent voice for each character.