Use Talking Photo when you have a still portrait and want the person to speak or sing. If your source is already a video, use Lip Sync instead.
Create your talking photo
Open Talking Photo and sign in.
Select Add Image and choose a clear portrait with one visible face.
Select Add Audio. Upload your recording, choose a preset, use Text to Speech, or use the voice-cloning option shown.
Choose Realistic or Expressive, then select the section of audio you want to use.
Choose a supported resolution. In Expressive mode, optionally describe the movement or expression.
Review the credit estimate and select Generate Video.
When the result completes, review it in History or My Library and download it.
Which mode should I choose?
Mode | Best starting point | Current clip limit |
Realistic | Lip sync and natural facial detail | 0.5–300 seconds (up to 5 minutes) |
Expressive | Creative motion guided by an optional prompt | 0.5–45 seconds |
The selected audio must fit the mode. If a longer recording is rejected in Expressive, trim it to 45 seconds or select Realistic when that suits the result. These are generation limits; they are separate from how long a job takes to process. Follow the current editor if its options have changed.
Use your own words, voice, or language
An uploaded recording supplies the speech or singing. Talking Photo animates the face to that audio; it does not translate it. For example, upload a Spanish MP3 to make the portrait speak that Spanish recording. To create a new voiceover from a script first, use the voice generation guide.
Improve mouth movement and facial detail
Keep the face clearly visible, well lit, and unobstructed.
Use clear audio and confirm the selected section contains the speech you want.
Start with Realistic when lip-sync accuracy is the main goal.
Test a short section before a longer generation. A prompt or higher resolution cannot guarantee exact facial identity or perfect mouth movement.
If the result is silent, cut short, or technically broken
Check the selected audio section, player mute and volume, and whether the downloaded file has the same problem. A completed but broken result needs project review. Stop repeating the same request and send support the project link, mode, resolution, selected audio length, visible error, and the time in the clip where the problem occurs. Keep the project until it has been reviewed. See generation failures and credits.