To lip-sync a Suno music video, give a suitable lip-sync tool the exact audio section and the face or performance shot it should animate. Review a short result, place it at the matching song timestamp, and keep the final song recording fixed while you edit. Common problems include mismatched crops, an unsuitable face shot, or a later change to the audio.
You do not need a visible singer throughout the video. Choose the moments where a performance adds something, then use other shots where seeing a mouth would create unnecessary work. This guide explains a workflow using separate lip-sync tools and a video editor; it does not describe a new lip-sync feature inside LyricVideoMaker.
Decide whether you need mouth animation or timed lyrics
Lip sync changes a visible character's mouth to match the vocal. Lyric synchronization controls when words appear on screen. A caption can arrive at the right moment while the character sings the wrong syllable; those are separate checks.
Use that distinction to plan your shots:
| What the viewer sees | What needs to match the song |
|---|---|
| A clear, close view of the singer | Mouth movements, phrasing, and the chosen audio section |
| A lyric-led scene without a performing face | Text timing, readability, and visual pacing |
| A landscape, object, or abstract sequence | The edit's rhythm and mood; mouth animation is irrelevant |
| Two people with visible faces | Which person is supposed to perform each part |
For a first version, identify one useful sung phrase rather than processing every shot. A close-up on the hook may carry the idea while an instrumental passage works better without a face at all.
Choose a tool that matches your visual input
If you already have performance footage, investigate a video-and-audio lip-sync workflow. Sync's documentation describes this input pairing, including rules for different audio and video lengths. If you start with a portrait, choose a model that explicitly accepts a still image and audio; Hedra Avatar lists that workflow and supports vocal performances.
These inputs are not interchangeable. A still-image avatar model must create facial motion from the portrait, while a video-based process has an existing performance to work with. Check the selected model's current input requirements before producing a batch.
Start with your own performance or a character and materials you have permission to use. A model accepting a face or a recording does not establish permission to impersonate the person represented.
For an image-driven test, Hedra recommends a well-lit, front-facing portrait. Begin with a simple composition and a clearly visible mouth before attempting more difficult framing. Portrait guidance.
Avoid designing the entire sequence around a microphone directly covering the lips. You can keep the microphone in the scene while choosing a camera position that leaves the performance readable.
Prepare a vocal guide without losing its timing
Keep the finished full mix as your timing reference. Save its filename and do not replace it midway through the edit.
For song lip sync, Sync recommends isolating the vocals instead of submitting a mix dominated by backing music. Its guidance also favors one speaker per audio input. That makes the lead vocal a useful guide when the arrangement or backing singers compete with it. Audio input guidance.
Listen to the isolated track before using it. Missing consonants, a cut-off word, or leftover backing vocals can make it a poor guide. A cleaner-sounding file is not automatically better if the process has removed part of the performance.
Most importantly, preserve the stem's position relative to the full mix. If you remove silence at the beginning of the vocal file, its first sample no longer represents the same point in the song. Either retain that silence or record the offset and apply it deliberately when preparing each shot.
Keep the guide and the listening mix separate:
- Vocal guide: used to drive the mouth animation where the selected tool recommends it.
- Full song mix: used as the final soundtrack in the assembled video.
This lets you review the visual result against the real recording without accidentally replacing the song with the isolated vocal.
Cut matching audio for each performance shot
Use actual timestamps rather than labels such as “second chorus.” A repeated chorus can have different phrasing, extra words, or a longer final note.
Here is an illustrative edit, not a measured generation result:
| Item | Example |
|---|---|
| Phrase starts in the full song | 41.20 seconds |
| Phrase ends in the full song | 46.60 seconds |
| Audio supplied for this shot | The interval from 41.20 to 46.60 seconds |
| Requested performance duration | 5.40 seconds |
| Where the returned shot begins on the master timeline | 41.20 seconds, if no extra lead-in was supplied |
If you add a lead-in to protect the start of a word, include it in both your crop notes and placement calculation. For example, a crop beginning at 40.90 seconds has 0.30 seconds before the phrase; place the full returned clip at 40.90, not 41.20.
For video input, inspect how the tool handles a duration mismatch. Sync documents modes that trim, loop, reverse-repeat, or change the video's speed to fit the audio. Those choices can visibly change a performance, so match the input durations where practical rather than allowing an unexpected default to decide the edit. Sync mode behavior.
For image-driven generation, use the selected model's audio-duration rules. Do not assume a duration setting from a video-based tool also applies to a still portrait.
Review one phrase before processing the rest
Choose a phrase with a clear beginning and a sustained or emphasized syllable. Review the same short section several ways:
- At normal playback speed: does the face feel like it is performing the song?
- Around the first clear word: does the mouth action start with the right part of the vocal?
- Near the end: does alignment stay consistent through the phrase?
- With the full mix restored: does the result still work in the listening context viewers will hear?
- At the intended crop: can the face remain clear in a vertical or horizontal edit?
Then classify the problem before deciding to regenerate:
| Symptom | First thing to inspect | Practical next action |
|---|---|---|
| Mouth movement is late by about the same amount throughout | Crop start and clip placement | Correct a constant picture offset within the available edit handles |
| The opening fits but the ending drifts | Duration, speed changes, or a different audio rendition | Recheck the exact source and timing before generating again |
| Timing feels right but the mouth forms the wrong shapes | Vocal guide and the model's interpretation | Try a clearer guide or a different suitable shot/model |
| Another person's mouth is moving | Target face selection | Select the intended performer or simplify the shot |
| A face or mouth becomes distorted | Obstructions, angle, lighting, and input quality | Improve the source or replace the difficult shot |
At 30 frames per second, three frames equal 0.10 seconds. That can help describe a consistent offset, but moving a clip will not repair incorrect mouth shapes. If you need to move the picture earlier or later, keep the master audio fixed and check both neighboring cuts for gaps or lost action.
Assemble the performance and supporting shots
Put the accepted clips into your video editor at their recorded song positions. Use one continuous full mix for listening and mute duplicate clip audio where appropriate, so the guide tracks do not double the vocal or remove the instruments.
Check every join after assembly. A shot can work in isolation but feel wrong when it starts halfway through a word or cuts away before a phrase finishes. Where the performance does not earn a close-up, use a supporting shot instead of forcing a weak mouth animation into the edit.
Hedra's music-video guidance similarly treats the project as planning, focused clip generation, continuity review, and assembly. The final edit still needs a deliberate relationship to the song. Music-video workflow.
Keep mouth animation ahead of final decorative overlays where your workflow allows. Large text or effects covering the lower face make it harder to judge the result. Add the finished typography once the performance and timing have been reviewed.
When a lyric video is the more useful result
If the main goal is helping listeners follow the words, a visible singing character may not be necessary. LyricVideoMaker's audio-to-lyric-video workflow is an entry point for an existing song, and the lyric video guide compares production approaches.
If you want imagery around the music, review the AI music video page. The lip-sync stage described in this article uses separate tools; synchronized captions or generated scenes should not be mistaken for verified singing-mouth animation.
Whichever route you choose, finish with the actual exported file. Watch the first sung line, the main hook, and the ending with the soundtrack enabled. If you change the recording afterward, recheck both mouth movement and lyric timing.
Common questions about Suno music video lip sync
Do I need to lip-sync the whole song?
Only shots with a visible performing face need that treatment. Build the edit around what helps the song: performance, narrative, abstract visuals, or a mixture. Instrumental passages do not need mouth animation merely to fill time.
Can I fix every error by shifting the video?
A constant offset may be an editing problem. Drift or incorrect mouth shapes needs a different diagnosis. Check the inputs and any timing transformations before spending another generation on the same pair of files.
Should I upload the full mix or an isolated vocal?
Follow the selected model's current guidance. Sync explicitly recommends a vocals-only track for songs. Preserve its alignment with the full mix, and use the full recording again when reviewing the assembled music video.
