The best AI models for lip sync
Make a face speak or sing the audio you have. These are Call Sheet's picks for this job, in order, with the reason for each. Call Sheet is a free tool by Melissa Beery that ranks 101 image, video and sound models for a creative job. Its data was last updated on 20 October 2026.
1
Sync lipsync‑2 / sync‑3
Sync Labs
Built only for this. Any language, any footage, priced per second, and sync‑3 handles profiles and hands.
Replaces the mouth in footage you already have to match new audio, in any language.
- Cost
- Hobbyist US$5 plus US$0.05 per second, watermarked. Creator US$19, no watermark. sync‑3 up to US$0.133 per second
- Where
- Higgsfield
- Rights
- No restriction on commercial output found.
- Watch out
- A static AI face gives static lips, so generate the source already speaking.
How to ask itFront-facing footage already speaking + one clean voice track + model pick (lipsync-2 for most, sync-3 for profiles, close-ups and obstructions)Face frontal or near frontal, well lit, at least 480p, and large in frame.
One speaker per audio track. Use Segments for a second speaker.
Open Sync lipsync-2 / sync-3 in Call Sheet →
2
Kling Lip Sync
Kuaishou (Kling)
Lip sync on a Kling clip inside the same app, 60 seconds at 1080p through the API.
Adds mouth movement to a Kling clip from an audio file or a typed line.
- Cost
- Standard US$6.99 promo, 660 credits, 1080p, no watermark
- Where
- Higgsfield · check in app: OpenArt
- Rights
- Paid tiers are marked for commercial use.
- Watch out
- Human faces only, no animals.
How to ask itA Kling clip with a complete human face + uploaded voice or Kling TTS at 0.8x to 2x + (API) face_id and sound_insert_timeThe face must be complete and human. Animals are refused.
Keep the clip within the audio length, or expect the audio cropped.
Open Kling Lip Sync in Call Sheet →
3
Higgsfield Lipsync Studio
Higgsfield
Eight engines in one room, on the subscription you already have.
Eight lip sync engines in one room with scene templates.
- Cost
- Starter about US$9, 120 credits. Plus US$49, 1,000 credits
- Where
- Higgsfield
- Rights
- No restriction on commercial output and no separate licence to buy.
- Watch out
- A carrier, so quality and ceiling change with the engine.
How to ask itSharp front-facing source image or clip + expressive audio generated first + engine pick + scene template (General, Selfie, Podcast, Car Talking)Source in sharp focus, even light, front‑facing, no heavy makeup or glasses.
Give the source some expression. A flat face animates flat.
Open Higgsfield Lipsync Studio in Call Sheet →
4
Runway Act‑Two
Runway
Act the line yourself on a phone and the character's face follows. A performance, not just a mouth.
Act the line on a phone.
- Cost
- Standard US$15 (US$12 yearly), 625 credits. 5 credits per second, 3 second minimum
- Where
- Runway
- Rights
- Runway claims no ownership of your content.
- Watch out
- 30 seconds, 720p class output, and a real performance recorded per line.
How to ask itDriving performance video (one person, waist up, face visible, no cuts, under 30 seconds) + character image or video + gesture control on for imagesOne subject, framed no wider than the waist, face visible the whole take.
No cuts in the performance video.
Open Runway Act-Two in Call Sheet →
5
HeyGen Avatar IV & Lipsync
HeyGen
Photo plus script to a 30 minute presenter, with the strictest written consent rule here.
A photo plus a script becomes a presenter with gestures.
- Cost
- Creator US$29, 600 credits
- Where
- HeyGen's own site
- Rights
- You own output on paid plans and commercial use is allowed.
- Watch out
- Credits burn fast and it is an avatar product first.
How to ask itOne face per photo + script in any of 175 plus languages or an audio file + Speed for a front-facing single speaker, Precision for profiles, hands and groupsGet explicit consent from any real person before building a custom avatar.
Clear audio and a visible face, or accuracy drops.
Open HeyGen Avatar IV & Lipsync in Call Sheet →
6
OmniHuman 1.5 (Dreamina)
ByteDance
The most expressive photo‑to‑performance model, on terms written for private use.
One photo plus a voice becomes a performance that reads the emotion in the audio and moves the whole body.
- Cost
- Free daily credits. Paid Dreamina Basic exists, price not shown
- Where
- ByteDance's own site
- Rights
- Dreamina's terms say the service is for private, non‑commercial use unless additional terms allow it.
- Watch out
- Private‑use terms and a reported forced watermark on paid output.
How to ask itFront-facing well-lit photo + uploaded audio or a Dreamina voice + line and action description under 240 characters each + Avatar Pro or TurboNeutral, front‑facing, evenly lit face. No sharp angles or covered features.
Match the voice's tone and pitch to the face.
Open OmniHuman 1.5 (Dreamina) in Call Sheet →
7
LivePortrait
Kuaishou (open weights)
Free and local. Drives a face from a video you shot, no audio.
Free and local.
- Cost
- Free, your own GPU
- Where
- Kuaishou (open weights)'s own site
- Rights
- MIT code.
- Watch out
- No audio drive.
How to ask itSource portrait or portrait video + driving video of your own face acting the line (or a saved .pkl motion template) + flags for stitching, eye and lip retargetingKeep head movement modest. The warp breaks on big turns.
Match framing between source and driver as closely as you can.
Open LivePortrait in Call Sheet →