How to Transcribe Audio to Text in After Effects
Can After Effects transcribe audio? At the time of writing, After Effects has no built-in speech-to-text tool that turns dialogue into text layers. Captyne adds one: it transcribes your comp or a media file with word-level timing, right inside the After Effects panel, and keeps the transcript with your project.
What After Effects can and can't do on its own
| Task | Native After Effects | With Captyne |
|---|---|---|
| Show an audio waveform | Yes | Yes |
| Turn speech into text | No built-in transcription | Yes, with word-level timing |
| Time text to each spoken word | By hand, with keyframes or markers | Automatic |
| Build animated caption layers | By hand | One click with ▶ Generate Captions |
Without a plugin, the usual workaround is to transcribe in another app and retype or paste the text into text layers, then time every line yourself. That works for a title or two, but not for a full video.
Step 1: Choose the audio source
Open your comp, then open Window → Extensions → Captyne. On the Setup tab, under Source, set Source Type:
- Active AE Comp: transcribes the audio of the comp that's open. This is the usual choice.
- Manual Media File: click Choose Media File and pick an audio or video file, for example a voice-only stem.
Tip: mute music and sound-effect layers before transcribing from Active AE Comp. Clean dialogue gives the best results.
Step 2: Pick a provider
Choose a Provider on the Setup tab. Every option returns word-level timing, so your captions animate the same way whichever you use.
| Provider | Runs on | Cost | Setup |
|---|---|---|---|
| Offline / Local (Free) | Your computer | Free and unlimited on every plan | Download the ~3 GB Whisper model once. Needs an NVIDIA RTX 20 / GTX 16-series or newer GPU on Windows, or Apple silicon on Mac |
| Managed API — Deepgram Nova-3 | Captyne's servers | Uses your plan's monthly minutes | None. Works on any computer |
| OpenAI — Whisper-1, Groq — Whisper-large-v3, AssemblyAI — Universal-2 | That service | Billed by that service | Your own API key under ⚙ Settings → Providers |
With Offline / Local, your audio never leaves your computer. Cloud providers return normal capitalisation and punctuation, and write numbers as digits ("20%"). For a detailed comparison, see Offline vs cloud transcription.
Step 3: Set the language and options
Under Options:
- Language: leave it on Auto-detect for most clips. If the clip starts with music or silence and the text comes out in the wrong language, choose the language yourself. Whisper supports 99 languages. If yours isn't listed, use Custom Language Code.
- Speakers: switch On for interviews and podcasts with more than one voice.
- Model: Default, Best for difficult audio, or Fast for quick drafts.
To transcribe only part of a long comp, set Transcribe to Work area only (uses fewer minutes). For final captions, use Whole comp. See Work Area Only, punctuation switches and cancelling.
Step 4: Click Transcribe
Click Transcribe. Captyne extracts the audio, sends it to the provider you chose (or runs it locally) and returns every word with its start and end time.
If you use Offline / Local for the first time and the model isn't installed, Captyne shows Whisper Model Not Found with a ⬇ Download Now (~3 GB) button. See Whisper model download failed if the download stalls.
Step 5: Review the transcript
With Open after transcribing switched on, Review Transcript opens as soon as the words arrive.
- Click a word to hear it, or press Space to play and pause.
- Double-click a word to fix it. Type two words to split it, or clear it to delete it.
- Click Looks good — go to Preview → when you're done.
Full details: Review Transcript: listen word by word and fix mistakes.
Where the transcript is saved and reused
You don't need to export the transcript to keep it.
- Saved with the project. The transcript and every edit you make are stored next to your
.aep, in a folder named after it (MyProject.captyne/), with one file per comp. Reopen the project and the panel restores it. See Your captions are saved with your project. - Transcript cache. If you transcribe the same audio again with the same provider and settings, Captyne reuses the earlier result instantly and no minutes are used.
- Move it elsewhere. On the Preview tab, open Transcript & Raw Data and click ↑ Export Session to save a
.jsonfile, then ↓ Import Session in another project.
Troubleshooting
- Offline / Local isn't available: your GPU isn't supported. See Offline transcription GPU requirements and use the Managed API or your own key instead.
- Many words are wrong: see How to get more accurate transcriptions.
- You already have a caption file: skip transcription and use Import SRT / VTT. See How to import SRT or VTT captions.