Convert audio to text locally on your device for private, offline transcription with no API costs
Local Transcription processes audio files directly on your device without sending data to external servers. Using on-device AI models, it converts speech to text while keeping your data completely private and eliminating per-use costs. Perfect for sensitive content, offline use, or high-volume transcription needs.
Transcribe confidential meetings, legal recordings, or medical notes without cloud exposure
Capture interviews or field notes in remote locations without internet access
Process hours of content without worrying about API usage limits or costs
Convert voice memos and personal recordings to text privately
First time only: download the language model for your target language (100MB-1.5GB depending on size).
Models are cached locally and reused for all future transcriptions.
Provide an audio file or start a live recording through your device microphone.
Choose the source language and model size (Small for speed, Large for accuracy).
Auto-detect is available but may reduce accuracy.
Transcription runs locally and completes in seconds to minutes depending on file length and model size.
Transcribe this meeting recording locally [audio file]
Processing with Whisper Large locally... Complete! [00:00] Alice: Let's start with Q1 results. [00:12] Bob: Revenue is up 15%...
Convert my voice memo to text, keep it private
Using local model (no cloud upload). Transcribed: "Remember to follow up with the client about contract renewal. Priority: high..."
Transcribe this 2-hour interview offline
Started local processing with Medium model. Estimated time: 8 minutes. Progress: 25%... 50%... Complete!
| Parameter | Limit | Notes |
|---|---|---|
| File size | Unlimited | limited only by device storage |
| Duration | Unlimited | longer files take more processing time |
| Concurrent jobs | 1 at a time | queue additional files automatically |
| Model storage | ~1.5 GB max | per language model downloaded |
| Model Size | Speed | Accuracy | RAM Usage | Best For |
|---|---|---|---|---|
| Tiny | Very Fast | Good | ~1 GB | quick drafts, casual notes |
| Small | Fast | Better | ~2 GB | general use, meetings |
| Medium | Medium | High | ~5 GB | professional transcription |
| Large | Slow | Highest | ~10 GB | critical accuracy, difficult audio |
| Error | Meaning | Action |
|---|---|---|
| OUT_OF_MEMORY | Insufficient RAM for model | Use a smaller model or close other apps |
| MODEL_NOT_FOUND | Language model not downloaded | Download model from Settings → Models |
| UNSUPPORTED_FORMAT | Audio format not recognized | Convert to MP3, WAV, or M4A |
| PROCESSING_SLOW | Transcription taking too long | Use smaller model or upgrade hardware |