Most transcription apps require internet. You upload your lecture to a server, wait for processing, and hope the cloud service handles your data responsibly. With 5cut, you can select a downloaded on-device engine for offline transcription. Availability depends on the engine, language, device, and initial model download.
How offline transcription works
5cut uses on-device AI models that run directly on your iPhone's Neural Engine. The first time you select a language, the model downloads (typically 40-600 MB depending on the engine). After that, transcription works in airplane mode, on the subway, in a lecture hall with terrible WiFi — anywhere.
Four transcription engines
5cut offers multiple AI engines so you can choose the right balance of speed, accuracy, and model size:
| Engine | Languages | Model size | Best for |
|---|
| Apple Speech | 40+ | Built-in | Quick transcription, broadest language support |
| WhisperKit | 30+ | 39 MB – 1.5 GB | High accuracy, word-level timestamps |
| Parakeet | 25 European | ~200 MB | European languages, auto-detection |
| Qwen3-ASR | 30+ | ~700 MB | Multilingual, large vocabulary |
Why offline matters
- No 5cut cloud copy — 5cut has no server; choose a downloaded on-device engine for offline transcription
- No data caps — transcribe hours of recordings without eating into your mobile data plan
- Works everywhere — campus basements, trains, planes, libraries with blocked WiFi
- No per-minute costs — cloud transcription services charge per minute. On-device is free after the model download
- Speed — no upload/download wait. Transcription starts immediately
Supported languages
Between all four engines, 5cut supports transcription in over 40 languages. The exact list depends on which engine you choose and the language models available for your device.
Beyond transcription: remove silence too
5cut is primarily a silence removal tool. The typical workflow is:
- Import or record a lecture
- Remove detected silence automatically
- Transcribe the condensed version offline
- Export: video with burned-in subtitles, SRT file, or plain text transcript
Silence removal also works fully offline — it's a waveform analysis that never needs internet.
Speaker identification
5cut can identify up to 4 speakers in a recording and color-code them in the transcript. This works offline too. Useful for lectures with Q&A, seminars, or group presentations.
Free to start
5cut is free to download and includes in-app recording, silence removal, and transcription previews (first 5 minutes). Upgrade to Premium (with a 3-day free trial available) to unlock unlimited exports, full-length transcripts, speaker identification, and batch processing.
