What is the difference between lip sync and a talking photo in Craze?
Craze has two tasks for making someone speak, both in the Video tab:
- Lip sync starts from a video of a person and changes their mouth to match new speech. Use it to fix dialogue, swap a line or match a dub.
- Talking photo starts from a single picture of a person and turns it into a video of them saying the speech. Use it for presenters, avatars and characters.
Both need a speech clip. If you don't have one, make it first with the Voice task in the Craze Audio tab; see voiceovers.

How do you lip sync a video in Craze?
To lip sync a video to new speech in Craze:
- Open the Video tab on the Craze home page, or click Make in a project and switch to Video.
- Set Task to Lip sync.
- Add the Video of the person speaking.
- Add the Speech audio.
- Click Generate. The button shows the price.
The result appears in the project's Gallery. Models such as Sync Lipsync 3, Kling LipSync and LatentSync do lip sync, and each takes a different length of video and audio; the Video and Speech boxes show the limits for the chosen model.
How do you make a talking photo in Craze?
To turn a portrait into a talking video in Craze:
- In the Craze Video tab, set Task to Talking photo.
- Add the Portrait, a clear picture of the face.
- Add the Speech audio.
- Optionally write a Direction for the performance, for example "warm, slight smile, small nods".
- Click Generate.
Models such as OmniHuman 2, Kling Avatar 2 Pro, VEED Fabric 1.0 and HeyGen Avatar 6 make talking photos. The longest speech each one takes differs, from about a minute to several minutes, and the Speech box shows the limit for the chosen model.
How do you start a lip sync or talking photo from the Craze Gallery?
The Craze Gallery starts both tasks from the media you already have. Open the More menu on an item and look under "Make something from this":
| Gallery item in Craze | Next step | What it opens |
|---|---|---|
| Image | Make it talk | Talking photo, with the image as the Portrait |
| Video | Lip sync | Lip sync, with the clip as the Video |
| Audio | Lip sync onto a video | Lip sync, with the audio as the Speech |
| Audio | Voice a photo | Talking photo, with the audio as the Speech |
Add the missing piece and click Generate.
How do you get good results from Craze lip sync and talking photos?
Craze lip sync and talking photos work best with clean inputs:
- A clear face. Use a picture or video where the face is well lit, facing the camera and not covered.
- Clean speech. One voice, no music under it. Add music afterwards on the timeline.
- The right length. Keep the speech within the limit shown in the Speech box for the chosen model.
If the first take is off, click Reuse on the result in the Gallery to load the same settings, change the model or direction, and generate again.
Can the Craze agent make a talking presenter for you?
Yes. The Craze agent can write the script, generate the voiceover and make a talking photo or lip sync in one go. Ask it in the chat, for example "make my headshot present this script in a friendly voice", and attach the photo. The Craze agent shows you the models and prices before it generates. See the Craze models page for every lip sync and talking photo model.