Back to Skills

Local Transcription

Convert audio to text locally on your device for private, offline transcription with no API costs

AI ServiceComing Soon

What It Does

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.

On-Device ProcessingComplete PrivacyOffline CapableNo API CostsMulti-language

In a Nutshell

🔒
Private & Secure all processing happens locally, no data leaves your device
Offline Ready works without internet connection once models are downloaded
💰
Zero API Costs unlimited transcriptions with no per-minute charges
🎯
Fast Processing real-time or faster transcription on modern hardware
🌍
Multi-language supports 50+ languages with downloadable models

Use Cases

Sensitive Content

Transcribe confidential meetings, legal recordings, or medical notes without cloud exposure

Offline Fieldwork

Capture interviews or field notes in remote locations without internet access

High-Volume Work

Process hours of content without worrying about API usage limits or costs

Personal Journaling

Convert voice memos and personal recordings to text privately

How to Use

Step 1

Download the model

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.

Step 2

Upload or record audio

Provide an audio file or start a live recording through your device microphone.

Step 3

Select language and model

Choose the source language and model size (Small for speed, Large for accuracy).

Auto-detect is available but may reduce accuracy.

Step 4

Process and review

Transcription runs locally and completes in seconds to minutes depending on file length and model size.

Command Examples

You say:

Transcribe this meeting recording locally [audio file]

Assistant responds:

Processing with Whisper Large locally... Complete! [00:00] Alice: Let's start with Q1 results. [00:12] Bob: Revenue is up 15%...

You say:

Convert my voice memo to text, keep it private

Assistant responds:

Using local model (no cloud upload). Transcribed: "Remember to follow up with the client about contract renewal. Priority: high..."

You say:

Transcribe this 2-hour interview offline

Assistant responds:

Started local processing with Medium model. Estimated time: 8 minutes. Progress: 25%... 50%... Complete!

Limits & Behavior

ParameterLimitNotes
File sizeUnlimitedlimited only by device storage
DurationUnlimitedlonger files take more processing time
Concurrent jobs1 at a timequeue additional files automatically
Model storage~1.5 GB maxper language model downloaded

Models & Modes

Model SizeSpeedAccuracyRAM UsageBest For
TinyVery FastGood~1 GBquick drafts, casual notes
SmallFastBetter~2 GBgeneral use, meetings
MediumMediumHigh~5 GBprofessional transcription
LargeSlowHighest~10 GBcritical accuracy, difficult audio

FAQ

Setup Requirements

Sufficient disk space for models (75MB-1.5GB)
Minimum 4GB RAM (8GB+ recommended)
First-time model download (internet required)
Audio input capability for live recording

Troubleshooting

ErrorMeaningAction
OUT_OF_MEMORYInsufficient RAM for modelUse a smaller model or close other apps
MODEL_NOT_FOUNDLanguage model not downloadedDownload model from Settings → Models
UNSUPPORTED_FORMATAudio format not recognizedConvert to MP3, WAV, or M4A
PROCESSING_SLOWTranscription taking too longUse smaller model or upgrade hardware