The page could not load. Please refresh to try again.
This is taking longer than expected. You can keep waiting or refresh the page.
YOUR FACE. YOUR AUDIO.
Turn a portrait into a talking video, or match a face video to a new recording. Bring your image or footage, add your audio, and choose a compatible model in MagicCanvas.
Create a lip sync video →Opens Lip Sync in Create. Choose your model and review the credit cost before generating.
Find the right workflow ↓
START WITH WHAT YOU HAVE
PHOTO + AUDIO
Have a character image but no footage? Pair a clear portrait with your recorded narration. Choose an image-input model to create a talking video for an introduction, lesson, or character message.
Start with a photo →
VIDEO + AUDIO
Already have a face video? Choose a video-input model and add your replacement audio. Use a short section first, then review mouth timing and facial details before making more versions.
Start with a video →Open Lip Sync in Create. Pick a model that accepts your source: a still image or an existing video. The required inputs change with your selection.
Upload the portrait or footage and a separate audio recording. Start with one clearly visible face and one clean speaking voice.
Review the available duration and resolution settings, then check the displayed credit cost. Submit when the inputs are ready.
Play the finished video with sound. Check the beginning and end of sentences, pauses, mouth shapes, and facial details before downloading or sharing.
CHOOSE BY INPUT, THEN COMPARE
Available models and options can change. Create shows the current selection and price for your task.
| Model | You provide | Resolution | Audio limit | Get started |
|---|---|---|---|---|
| OmniHuman 1.0 | Photo + audio | Check in Create | Check in Create | Try model → |
| OmniHuman 1.5 | Photo + audio | Check in Create | Check in Create | Try model → |
| InfiniteTalk | Photo + audio | 480p / 720p | Up to 10 min | Try model → |
| InfiniteTalk (Video) | Video + audio | 480p / 720p | Up to 10 min | Try model → |
| MultiTalk | Photo + audio | Check in Create | Up to 10 min | Try model → |
For a photo, start with an image-input model. For existing footage, choose InfiniteTalk (Video). Compare a short recording first: a longer supported duration is a limit, not a recommendation or a guarantee of quality. If a model is unavailable, choose an alternative in Create.
Pair a character portrait with a short welcome or story introduction. Keep the first recording brief so you can evaluate how the face moves before building a longer scene.
Use your own narration for a presenter-style introduction or a short explanation. Review technical terms in the audio before generating, then add captions in your usual editor.
Prepare a translated voiceover separately and pair it with compatible face footage. Lip Sync follows the audio you supply; it does not translate the recording for you.
Try a new line or message with existing presenter footage using a video-input model. Compare the generated performance with the original before publishing the update.
Record different openings or calls to action for the same character. Generate and review each version, keeping the source material and recording quality consistent.
Give a portrait a spoken line from your story. For a full scene with camera movement and action, start with Image to Video or Text to Video instead.
Results vary with the source and model. Review the complete video for timing and visual artifacts.
VOICE OR A WHOLE NEW SCENE?
Use Lip Sync when you already have the audio you want your character to speak. For new visuals, start with a photo or a written scene instead.
AI lip sync creates mouth movements that follow an audio recording. In MagicCanvas, start with a portrait or a face video, choose a compatible model, and add the audio you want the person to speak.
Yes. Choose a model that accepts an image and audio, such as an available OmniHuman or InfiniteTalk image model. Upload a clear portrait and your recording. The result is a video, not an edited still image.
Choose a video-input model, such as InfiniteTalk (Video), when it is available in Create. Add your face video and a separate audio track. Image-input models require a portrait instead.
Yes. Upload a recording you have made. Record one speaker clearly, reduce background noise, and trim unwanted silence before uploading. Lip Sync follows the supplied recording; it does not clone a voice from the portrait.
No. Prepare the spoken audio in your target language first, then use it for lip sync. Check pronunciation, mouth timing, and the complete result before publishing a dubbed version.
This workflow uses an uploaded audio track. Record your script or create a voiceover separately, then add that audio in Create. A text prompt alone does not replace the required audio input.
Start with the input you have: a photo or a video. OmniHuman and MultiTalk configurations accept a photo with audio; InfiniteTalk has image and video variants. Create shows the currently available models, required inputs, settings, and cost.
Generation uses credits or diamonds. The required amount depends on the model, duration, and applicable settings. Check the displayed cost in Create before submitting a task.
A clear, unobstructed face and clean audio are good starting points. Strong profile angles, hands covering the mouth, fast head turns, and noisy recordings can make synchronization harder. Review facial details and timing throughout the output.
Limits vary by model. Check the selected model’s audio duration and output settings in Create before preparing a long recording. Start with a short test to evaluate the face and voice combination.
Use images, footage, and recordings you own or have permission to use. Obtain the person’s consent before making them appear to say new words, and do not present a synthetic performance as an authentic recording.
Choose a model, add your audio, and bring your next spoken idea to the screen.
Create a lip sync video →