Amlistening is an offline transcription app. It records what you say and transcribes it on your laptop, with no network. Bring in a GPS trace, and each line shows where you said it. Every take is a folder of plain files you keep.
Words appear as you speak, in about 80 languages with the default model, on the laptop in front of you. Nothing waits on a connection, and nothing you record leaves it.
Give a project the keywords you listen for, and each one lights up in the transcript as you say it.
Words appear as you say them, written by the nemo-speech sidecar the app launches on this machine, not an API call.
Click a line’s time to play just that line, and check it against what was said. Double-click to correct it; the machine’s text stays beside yours.
Bring in a trace from a GPS logger, a watch, a phone app or a GoPro, and every line of the transcript gets the place where it was said. The trace stays on your laptop: the map draws with no map tiles, so no request carries where you were.
The map follows the take as it plays, and each line shows the place where it was said.
Import a GPX or NMEA file, or a GoPro video with its GPS, and the app lines the trace up with the audio by time.
Start, mark and stop by voice or by key with the window out of sight. If the laptop dies mid-take, you lose seconds, not the recording.
Start, stop and mark a take without reaching for the window.
The menu-bar icon, in the notification area on Windows, starts, pauses, marks and saves a take.
Audio is written to disk as it arrives, so a crash or a power cut costs seconds, not the take.
Bring in audio from a phone, find any take by name or tag, and export captions, JSON, plain text or map files. Every take is a folder you can copy.
Drop a WAV, MP3, M4A or MP4 from a phone, a camera or another recorder, and it becomes a take like one you recorded.
A take starts with no form to fill in. Rename and tag it from its row, even while it records.
SRT or WebVTT captions for a player, JSON with every timing, plain text to read and search, or GPX and GeoJSON for a map tool.
Download the app for your computer, then pick a model in Settings. It downloads once.
Press Record, a key, or say “start recording”. A take has no form to fill in first. The transcript fills in as you talk; mark the moments worth coming back to.
Play any line, correct what it misheard, and export SRT or WebVTT captions, JSON, plain text, or GPX and GeoJSON for a map tool.
Anything else is in the docs, or write to support@amlistening.app.
Not to record or transcribe. Capture, transcription, review and export need no network, at any tier. You need a connection for three things: on the first run, to download the transcription model; to sign in and activate a Pro licence, which then refreshes itself whenever a network is around; and to get updates.
More in the docsNo. Nothing you record is ever uploaded, and the app has no telemetry. There is no cloud mode to switch on by mistake: no provider to pick and no API key to paste. Your takes stay in a folder on your machine.
More in the docsYes. Drop a WAV, MP3, M4A or MP4 from a phone, a camera or another recorder on the recordings list, and it becomes a take with a transcript, made on this machine. Several files import one after another.
More in the docsOnly if you bring in a trace. The app does not record your position. Import a GPX or NMEA file from a GPS logger, a watch or a phone app, or import a GoPro video, whose GPS comes in with its sound. The app lines the trace up with the audio by time. Then every line of the transcript has a place, a map shows where you said it, and you can export it all as GPX or GeoJSON. Importing a trace is part of Pro. A trace never leaves your machine: the map draws with no map tiles, so no request carries where you were.
More in the docsAny input your computer can hear: built in, USB or a headset. To move around while you talk, a wireless kit keeps the cable out of the way: clip the transmitter on and plug its receiver into the laptop. DJI Mic, DJI Mic Mini and RØDE Wireless GO work well. They are examples, not requirements; Amlistening is not affiliated with DJI or RØDE.
More in the docsIt depends on the microphone, the room and the speaker, so there is no single number we can stand behind. Any line plays its own audio, so you can check it against what was said, and you can correct any line. The download is free: try it on your own voice first.
More in the docsAbout 80 with the default model, Nemotron 3.5. You choose the language for each project; it is never guessed.
More in the docsNo. Lines are not attributed to speakers. Each line is transcribed once, as you speak, or in the background for a file you import, and the model does its best in that one pass; nothing rewrites it afterwards, and there is no summary. You can correct any line yourself. For a polished version or a summary, export the take as plain text or JSON and give it to whichever AI agent you already use. That step is yours: the app never sends your transcript anywhere.
More in the docsNo. It records a microphone input, not the sound of other apps, and nothing joins your meeting. It is made for you speaking, not for capturing a call.
More in the docsEach is good at a different job. MacWhisper is a Mac app for files, meetings and dictation. Vibe is free and open source, and it transcribes audio and video files. Wispr Flow turns speech into polished writing in other apps, and it transcribes in the cloud. Amlistening records what you say into a microphone, on your own machine, and keeps each word as spoken. The comparison says where each one does more, from each product’s own pages.
Read the comparisonEach take is a folder inside your project: the audio as WAV chunks, and the transcript and markers as JSON lines. Copy the folder and the take goes with it.
Apple silicon Macs, and 64-bit Windows and Linux. Intel Macs are not supported.
More in the docsYes. Every update is free, at every tier: new features, fixes and, as they are added, new transcription models, since the list of models comes with the app. It checks for a new version at launch and downloads one only when you say so, or by itself if you turn on Update automatically.
More in the docsThe app falls back to Free. Nothing is locked: every take you recorded stays readable, playable, correctable and exportable.
More in the docsNo. The source is not published.
Nothing you record leaves your machine.
