How to Lip Sync a Video

Prepare existing footage and replacement speech, match their timing, and check mouth movements, face continuity and sound before keeping a dubbed take.

LumiYing TeamUpdated

Start with a line that fits the performance

To lip sync a video, choose a short clip with one clearly visible speaker and prepare the new line as MP3 or WAV. Open Lip Sync AI, choose an audio-and-video-compatible model and upload both. Output length follows the source automatically. Models that require a fixed length use the next supported duration that covers the whole clip.

A matching file length alone is not enough. A rushed sentence can collide with a long pause in the original performance. Watch the source on mute, mark where speech begins and ends, and record a line with a similar rhythm. Review the generated sound as well as the face because this workflow regenerates the performance.

Prepare, submit and judge the take

  1. Step Choose an uninterrupted shot

    Use one speaker facing forward or at a mild angle. Avoid a hand, microphone or object covering the lips. Cut a longer recording into a short, complete thought before uploading. Fast turns and edits make mouth changes harder to evaluate.

  2. Step Record against the source timing

    Watch the clip while rehearsing the replacement line. Keep a little silence before the first word and preserve natural breaks. Record one voice without loud music, clipping or overlapping speech. Export MP3 or WAV and listen from beginning to end.

  3. Step Make the recording fit

    Follow the selected model’s displayed media limits. The recording must fit inside the source video. For an 8-second video, prepare at most 8 seconds of speech. Shorten or re-record a line that does not fit.

  4. Step Upload both files and review the quote

    Open Lip Sync AI and choose a model. Add the source MP4 or MOV and replacement recording. Select resolution and frame where offered, then wait for media checks. Output length is determined from the source, rounded up to a supported duration when required. The quote follows the model’s source/output billing rules.

  5. Step Compare before keeping the result

    Play the full take at normal speed with sound. Compare the first word, consonants that close the lips, pauses and the last word. Inspect teeth, chin and face edges during movement, then compare hair, clothing and background with the source. Listen for the old line underneath the new voice. Use Re-edit to change the responsible input and compare again.

A practical preflight check

Before generating

Video
One speaker with a readable mouth
MP4/MOV. Depending on the model, up to 50 or 200 MB and 15 or 30 seconds. Keep the mouth clear.
Recording
A complete, clearly spoken phrase
MP3/WAV, up to 15 MB. Follow the model’s audio limits and fit the recording inside the source video.
Timing
Words and pauses fit the existing action
Keep the opening breath and final word inside the selected clip.
Picture
Automatic length, model-specific framing
Length follows the source. Fixed-duration models round up to cover the clip. MiniMax H3 offers frame choices.

Three comparisons to practice with

Listen to the replacement recording, then compare the source and example at normal speed.

A seated introduction

Start with the first word. A direct view makes early or late mouth movement easy to spot. In a quiet gap, check that the mouth returns to rest instead of continuing to speak.

Load this example
Source video
Replacement speech
Lip-synced example

Fix the input behind the mismatch

Speech is cut off
Audio duration and the complete final word
Shorten or re-record the phrase before uploading. Do not speed it up until pauses disappear.
Mouth keeps moving in silence
The original performance and new silent gaps
Use a line with closer pacing or a source with quieter movement. Review another take.
Words or voice change
The uploaded recording against the generated sound
Use a cleaner single-speaker recording. Replacing audio in an editor still requires checking lip timing.
Face or surroundings drift
Head turns, occlusion and source detail
Try a clearer, steadier source. A generative edit can alter details beyond the mouth.
Old speech remains
The final soundtrack with headphones
Reject the take if dialogue is doubled. Prepare a clean replacement and review the next result.
Frequently asked questions

Frequently asked questions

Does this automatically translate a video?

Prepare the spoken line in the target language yourself. This page uses the uploaded recording and does not provide transcription, translation or text-to-speech.

Can I choose one face in a conversation?

The first workflow is for a single visible speaker. Split a conversation into individual shots before preparing replacement speech.

Can I reuse the original soundtrack?

You can upload a separate MP3 or WAV, but a recording containing several voices or loud music makes the target speech less clear. Keep a copy of the original and compare the generated soundtrack.

Try a new line on your video