How to make a talking AI avatar with lip sync
Want a photo to talk? Dora's Lip Sync model animates a face from a still image and an audio track, matching mouth movement to speech or song. Here's how to make a convincing talking AI avatar.
What you'll need
- A clear, front-facing photo of the face.
- An audio file of the speech or singing (the words your avatar will "say").
Step 1 — Choose the Lip Sync model
On the create page, select Lip Sync from the model menu. It's built for exactly this: faces + audio → a talking clip.
Step 2 — Upload the photo
Use a sharp, well-lit, front-facing portrait. Avoid heavy occlusion (hands over the mouth, sunglasses) and extreme angles — the cleaner the face, the better the mouth tracking.
Step 3 — Add the audio
Upload your voice track. Clear audio with minimal background noise produces the most accurate lip movement. Keep clips reasonably short for the crispest sync.
Step 4 — Pick resolution and aspect ratio
Choose 9:16 for social or 16:9 for landscape, and a resolution that fits your use. Generate, then review the timing.
Tips for believable avatars
- Match the energy. Calm audio + a calm portrait reads as natural; mismatched energy looks off.
- Front-facing photos sync best.
- Short takes beat long monologues for accuracy — break long scripts into segments.
- Need a voice too? Generate one with Dora's ElevenLabs voice model, then feed it in.
Where this is useful
- Faceless or avatar-based content channels
- Localizing a message into multiple languages
- Bringing a portrait, mascot, or character to life
Related
To animate a photo without speech (just motion), use Gemini Omni photo-to-video instead.