AI video lip-sync test: who actually talks?
Winner: Veo 3.1 Fast. It was the only one of four models where the character visibly delivers her line. PixVerse V6 framed a good close two-shot but the line was hard to see being spoken, Kling spoke briefly and then froze, and Runway barely showed any talking.
| # | Model | Overall | Adherence | Motion | Quality | Artifacts | What we saw |
|---|---|---|---|---|---|---|---|
| 1 | Veo 3.1 Fast | 7.5 | 8 | 6 | 8 | 8 | Close two-shot, she clearly speaks; natural expressions. |
| 2 | PixVerse V6 | 7.0 | 7 | 6 | 8 | 7 | Close two-shot with natural expressions, but her line is hard to see being spoken. |
| 3 | Runway Gen-4.5 | 6.5 | 6 | 6 | 7 | 7 | Wide shot on bench, mostly smiling; little visible dialogue. |
| 4 | Kling 3.0 | 6.3 | 6 | 5 | 7 | 7 | She speaks briefly, then both freeze; stiff. |
adherence 8motion 6quality 8artifacts 8
Close two-shot, she clearly speaks; natural expressions.
adherence 7motion 6quality 8artifacts 7
Close two-shot with natural expressions, but her line is hard to see being spoken.
adherence 6motion 6quality 7artifacts 7
Wide shot on bench, mostly smiling; little visible dialogue.
adherence 6motion 5quality 7artifacts 7
She speaks briefly, then both freeze; stiff.
Tips for dialogue prompts (from this test)
- Ask for a close shot. Veo framed a close two-shot and the speech read clearly. Runway went wide, and the mouths were too small to judge.
- Keep the line short. A 4-second clip fits about one short sentence. Kling spoke briefly and then froze, so give the scene an action to fill the remaining time.
- Put the exact words in quotes and say who speaks ("she says ...").
- Use a model with native audio. Runway Gen-4.5 is silent, so you'd have to add a voice and sync it later.
Try Google Veo · All AI video generators with audio · AI video prompt guide · Best for people & faces
FAQ
Which AI video generator has the best lip sync?
In our October 2026 dialogue test, Veo 3.1 Fast (7.5/10). It was the only model where the character clearly says her line on camera, in a close two-shot with natural expressions.
Does Runway Gen-4.5 do lip sync?
Not natively. Gen-4.5 generates silent video, and in our test it showed a wide shot with little visible talking. Runway offers separate audio tools, but that is a different workflow from native dialogue generation.
Is this a full lip-sync benchmark?
No. It is one dialogue prompt with one generation per model, so it shows direction, not proof. Dedicated avatar tools such as HeyGen are built for talking-head lip sync and are a different category. We haven't scored them yet.
What prompt did you use?
"Two people having a conversation on a park bench; she says 'I told you it would work.'" It's 4 seconds at 720p, with audio on for Veo, Kling and PixVerse.