Record, transcribe, copy
Three things, in that order, on one screen. Speech recognition runs on the device, so the app behaves the same on a plane, in a basement, or with the radio switched off — there is no first-run download and no server behind it.
Tap once to record
Up to 30 minutes in a single take. The timer counts up while you talk; tap again to stop, and transcription starts on the spot.
Or open a file you already have
MP3, M4A/AAC, WAV, FLAC, OGG Vorbis and Opus. Files are chosen through Android's own document picker, so the app gets the one file you picked and nothing else — and needs no storage permission.
The model ships inside the app
OpenAI's Whisper, the ggml-base-q5_1 build, 57 MB, running on the phone's own processor. Nothing is uploaded and nothing is streamed.
One permission, the microphone
RECORD_AUDIO is the only permission the app requests. It holds no INTERNET permission, so it cannot open a network connection at all — Android enforces that, rather than us promising it.
Many languages, detected
The model handles a wide range of languages and works out which one it is hearing. There is nothing to choose before you start talking.
Copy, and that is the export
The transcript lives in the app while it is open — no save button, no export menu, no file to manage. Copy puts the text on the clipboard, and from there it goes wherever you paste it.
It stops when you leave
Android silences the microphone for an app that is not in the foreground. Rather than pretend to keep going, the app stops and transcribes what it captured up to that point.
Straight about long files
A very long recording may be refused, with a message saying roughly how long a file this phone can manage — better than a spinner that never finishes.
At a glance
What the app is, in the terms a store listing asks for.
- Publisher
- Methodox Technologies, Inc.
- Package name
io.methodox.voicetranscriber.mobile- Version
- 0.9 — first release
- Requires
- Android 7.0 (API 24) or newer, 64-bit
- Orientation
- Portrait, one screen
- Permissions requested
RECORD_AUDIO, and nothing else- Network access
- None. The app holds no
INTERNETpermission - Speech model
- Whisper
ggml-base-q5_1, 57 MB, bundled in the app and run on the device - Audio formats read
- MP3, M4A/AAC, WAV, FLAC, OGG Vorbis, Opus
- Recording length
- Up to 30 minutes per recording
- Account
- None. There is nothing to sign up for and nothing to sign in to
- Contact
- contact@methodox.io