remotion-narration
Skill0 starsChanged 2 days agoMITView on GitHub
Turn a video script into per-scene voiceover for Remotion in one command - one MP3 per scene in a saved or cloned voice, measured durations converted to durationInFrames, word-level captions in Remotion's Caption format, and an SRT. Use when building a narrated video with Remotion (or any React/code video tool), when the user asks to "add voiceover", "narrate these scenes", "sync scenes to the narration" or "add captions that follow the voice". Voice and timing by Miragea, no GPU or ElevenLabs account needed.
- Reads credentials
In this repository10 other files it gives its agents.
- llms.txtllms.txt
- audiogram-clipsSkill
- auto-brollSkill
- hook-variantsSkill
- screen-recording-voiceoverSkill
- shorts-voiceoverSkill
- ugc-video-adsSkill
- video-captionsSkill
- video-dubbingSkill
- video-to-shortsSkill
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 0 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 23
Make sure MIRAGEA_API_KEY is set (never print it).
- Very specific: line 27
Write the script. Each ## Heading (or blank-line paragraph) becomes one scene:
- Very specific: line 42
Use public/narration/narration.json in the composition (see examples/Narrated.tsx): totalFrames for durationInFrames, each scene's …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Remotion narration
Writes everything a Remotion composition needs to be driven by its voiceover: audio per scene, exact durations, start frames and captions. Scene lengths come from the real audio, so nothing drifts.
When to use
- The user has a script (Markdown, text or JSON) and a Remotion project and wants narration.
- Scenes should last exactly as long as their voiceover; captions should follow the voice word by word.
Steps
- Make sure
MIRAGEA_API_KEYis set (never print it). Pick the voice:--voice VOICE_ID- a voice saved in the user's Miragea account, or--ref-url https://...- a public clip of the voice to clone (with the user's consent), plus--ref-text "exact words in the clip"for the cheaper 1x rate.
Write the script. Each
## Heading(or blank-line paragraph) becomes one scene:## intro Meet Retake, the fastest way to fix a voiceover mistake. ## demo Type the right word. Retake re-voices the sentence and splices it in.Generate into the Remotion
public/folder:python3 scripts/narrate.py script.md --voice VOICE_ID --out public/narration --fps 30- Use
public/narration/narration.jsonin the composition (seeexamples/Narrated.tsx):totalFramesfordurationInFrames, each scene'sstartFrame/durationInFramesfor a<Sequence>,staticFile('narration/' + scene.file)for<Audio>, andscene.captions(RemotionCaption[]:text,startMs,endMs,timestampMs,confidence) with@remotion/captionscreateTikTokStyleCaptions()for animated captions. - Edit the script and re-run freely: unchanged scenes are reused from the manifest at no cost; if a run stops midway, finished scenes are kept and the next run resumes.
Options
| Flag | Default | Meaning |
--fps | 30 | Frame rate for frame maths |
--pad | 0.4 | Seconds of breathing room after each scene |
--captions miragea | yes | Times captions from the generated audio (accurate; +3 points/s) |
--captions estimate | Free: spreads words by length | |
--model | expressive | standard follows the reference voice more closely |
--speed | 1.0 | 0.25-4 |
Scenes longer than 2,000 characters are split at sentence ends and joined automatically.
Troubleshooting
choose a voice-> pass--voiceor--ref-url. Reference URLs must be public, under 1,024 chars and served directly (GitHub raw, S3/R2 public objects work; signed redirects may not).401/402/429-> key, balance, rate limit. Re-running is always safe.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

