Magic Hour lip sync: workflow, pricing & troubleshooting

Magic Hour publishes this guide about its own Lip Sync product. Treat it as product documentation rather than an independent review; verify the current workflow, limits, pricing, and output on a short permitted sample before committing a full project.

When Magic Hour Lip Sync fits the job

A practical dialogue-replacement workflow

Imagine an onboarding video says “Click the green button,” but the interface now uses a blue button. Replace the spoken instruction with “Click the blue button” and supply the new audio to Lip Sync. Review the screen recording too: synchronized speech will not update the button shown in the picture.

1. Trim a continuous shot containing the visible speaker and the line to replace.
2. Record or generate the new line at a pace that fits the scene.
3. Listen to the audio by itself and approve every word.
4. Upload the video and audio to Magic Hour Lip Sync.
5. Watch the generated result through the start, pause, and end of the sentence.
6. Put the approved clip back into the edit and update any captions or on-screen instructions.

For upload screenshots, follow how to lip sync a video.

What is free, and when do you need a paid plan?

The current Lip Sync product page offers three guest generations per day without signup, with a ten-second maximum. Free video results may include a watermark. Account credits, saved projects, rewards, higher limits, watermark-free output and commercial-use eligibility depend on the current account and plan.

Paid plans currently add credits, higher export limits, API access, commercial use and watermark-free exports under the applicable terms. Prices are localized and can change, so use the current pricing page for the visitor's region and billing period instead of relying on a copied dollar amount.

Try the sample before choosing a plan. Pay for the duration, mode, and export your project needs after the result meets your requirements.

Before regenerating: input, render or download problem?

First establish whether you have unsuitable inputs, an unfinished or failed job, an inaccessible download, or a completed video with a visible defect. Those need different next steps. A project ID, a successful request or a download button does not establish that the speech and face are acceptable.

The current Lip Sync mode and timing guide explains these input checks. In the editor, inspect the actual video and audio selections before submitting again. Keep the face and lips visible; do not assume a different soundtrack will fix an unsuitable source.

Check the existing API job before submitting again

For API projects, GET /v1/video-projects/{id} distinguishes queued or rendering work from complete output. When status is error, inspect the error details; canceled is a separate state. Read the message on request or account errors such as invalid_request, unauthorized or insufficient_credits instead of diagnosing them as mouth-quality defects.

If a completed project’s signed download has expired, follow the existing-output recovery instructions and GET that same available project for its current download URL. Check expires_at: the response may reuse a still-valid cached URL, so requesting details does not promise another full 24 hours. Recover the file before creating another generation.

Make the next comparison reproducible

For example, if the line stops early, first compare the chosen audio and video end points. Correct one demonstrated selection mismatch, then check the full exported sentence. This is an illustrative diagnostic procedure, not a tested repair or a guarantee.

Keep the route, selected mode, source versions, video/audio trim positions, project ID and exact mismatch timestamp. Compare the original downloaded result with any later edit. If support is needed, retain the project and provide that record; a screenshot alone cannot show which files or segment produced the result.

What to fix when the result looks wrong

Compare the exact source clip, supplied audio, original download and any later edited version. The same visible symptom can have several causes; the table below gives inspection starting points, not a diagnosis or a guaranteed repair.

Distinguish three timing patterns

Constant offset: check recognizable speech moments near the beginning, middle and end. If the mouth is consistently early or late, try one reversible timeline alignment while keeping the untouched download. Accept it only if all landmarks improve without cutting words.

Growing drift: if the start aligns but the end does not, one global offset cannot align both. Compare durations, selected segments and later edits. Restore the unedited download as a control before changing an input. A visible drift is not proof of a frame-rate or model problem.

Local mismatch: if one word or pause fails while the rest aligns, moving the whole soundtrack can damage correct sections. Inspect that audio phrase and neighboring source/result frames. Change one evidenced problem, then review timing, teeth, jaw and identity through the complete new export.

If the source mouth is hidden during a turn, inspect another shot or use an appropriate cutaway. Keep the approved meaning and speech intact. Clearer inputs can reduce difficulty, but no single source change or setting guarantees an acceptable result.

Do not keep regenerating the same unsuitable inputs indefinitely. A reshoot or conventional edit may be the better way to finish a difficult scene.

Using Lip Sync through the API

The current Lip Sync API tool guide documents per-rendered-frame charging: Lite and Standard use one credit per rendered frame and Pro uses two. A ten-second clip capped at 30 FPS is about 300 credits at the one-credit rate. The completed job's credits_charged field is the exact cost; the initial value can be an estimate based on the eventual rendered FPS.

Use the API reference for upload fields, segment selection, and frame-rate settings. Keep API billing calculations separate from the free guest website offer.

Before publishing a dubbed or edited video

Check the message, face, audio, captions, and any visible demonstration together. For localization, have a fluent speaker review the translation. Make sure your use of the person's likeness, voice, and footage is authorized; a paid plan does not supply those permissions for someone else's media.

If you are still choosing a platform, use the lip sync comparison. If your next step is simply trying a recording, open Lip Sync with one short clip and its approved audio.

Product, plan and API facts were rechecked against the Lip Sync product page, pricing page, tool guide and API reference on September 13, 2026.