Learn the documented Runway lip-sync workflows for animating a still image with uploaded dialogue, working from a visible face in video, or transferring a recorded performance with Act-Two.
Choose the Runway lip-sync workflow
To use Runway Lip Sync, first identify what material you have. Runway describes AI lip sync as matching mouth movements in a video to a separate audio track so that a person or character appears to speak the words in that track. Its guide documents three starting points: a text prompt, a still image with audio, or a recorded performance transferred to another character (Runway).
For a still character image and uploaded dialogue, use Add Dialogue. Runway says this app animates a still photo from an audio clip. For a performance-led result, use Act-Two: it transfers mouth movements, expressions, and timing from a recorded video performance to another character or image (Runway).
Keep those workflows distinct. Add Dialogue is the documented audio-plus-image route; Act-Two uses a performance video to guide more than the spoken words.

Prepare the character and dialogue
Runway recommends a front-facing, evenly lit face with a clearly visible mouth. It also identifies profile angles, extreme head turns, low resolution, and poor lighting as conditions associated with weaker results. Fast or overlapping speech can drift out of sync, while close shots can reveal artifacts around teeth and the tongue (Runway).
For the voice track, Runway recommends clear audio with little background noise because it synchronizes better than noisy material (Runway). As an editorial suggestion, listen for competing voices, background sound, or rushed delivery before uploading the track. If possible, choose a section where one person speaks at a measured pace.
You should also confirm that you have the right to use every face and voice involved. Runway directs users to obtain permission and disclose synthetic or dubbed media when viewers would expect a genuine recording (Runway).

Use Add Dialogue with a still character image
Runway describes the Add Dialogue workflow as placing a still photo and an audio clip into the app so the face is animated to speak. It presents this as the route for creating a talking avatar, spokesperson, or character when a face and voice track already exist (Runway).
A practical sequence supported by Runway’s guide is:
- Choose a still image with a front-facing, well-lit face and visible mouth.
- Prepare or record a clean dialogue track.
- Upload the inputs and select the audio-plus-image approach.
- Generate a few seconds as a short test.
- Review the result before committing to the full clip.
- Download the finished MP4 for editing or use in a larger project (Runway).
As an editorial suggestion, use the test to inspect mouth timing, facial stability, and visible tooth artifacts. If the result is weak, try clearer source material before extending the sequence.
Work with character video or Act-Two
For footage containing a face, Runway’s general lip-sync process begins with a video clip in which the face and mouth are visible, paired with clean audio. Its guide recommends generating a short section first and exporting the completed result as an MP4 (Runway).
Act-Two serves a different purpose. You perform the line on camera, then Act-Two uses that video input to drive the mouth movement, expression, emotion, and timing of a different character or image (Runway). Choose this route when the delivery of the performance matters alongside the dialogue.
If generated frames appear soft or choppy, Runway suggests processing the clip with upscaling or frame interpolation (Runway). As an editorial recommendation, preserve the original output as well as any processed version so you can compare whether the additional treatment improves the intended shot.
Conclusion
Use Add Dialogue when your inputs are a still character image and an audio clip. Use a visible-face video with a separate audio track for the documented video lip-sync approach, or choose Act-Two when a recorded performance should drive another character’s mouth movement, expression, and timing (Runway).
The most practical recommendation is to begin with clean audio and a well-lit, front-facing face, then generate only a few seconds. Runway specifically recommends a short test because it can reveal problems before you process a full clip (Runway). Proceed with the longer generation only after the test supports the result you want, and disclose the synthetic or dubbed presentation where viewers would expect authentic footage.
Sources
Disclosure
This article was prepared with AI assistance.
Improve this guide
Report a problem or suggest the next question
Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.
