Turn audio recordings into clean Markdown.
Meetings, interviews, voice memos and podcasts transcribed into a structured document — summary first, then a clean transcript with speakers.
Your files are never stored: documents convert in your browser where possible, and anything sent to our AI travels encrypted (HTTPS) and is discarded after processing. Privacy Policy
A raw transcription is a wall of text; CleanMD gives you a document. Upload an MP3, WAV, M4A or OGG (up to 25 MB) and the AI produces a short summary followed by a full transcript, broken into paragraphs at natural pauses, with ### sections for distinct topics and speaker labels when voices are distinguishable. Disfluencies are cleaned up without changing a single meaning. The file is processed only for the duration of the conversion — nothing is stored afterwards.
Working with related formats? See also Video to Markdown, YouTube to Markdown.
From a recording to notes you can search
A raw transcript is rarely what you want from a recording. Speech is full of false starts and repetitions, a meeting jumps between topics, and an hour of audio turned into one block of text is almost as hard to use as the audio itself.
CleanMD sends the file to an AI model and asks for a document rather than a transcript. You get a summary first, then the cleaned transcript split under headings by topic, with speakers labelled when their voices can be told apart, and with real names when they're mentioned in the recording. Filler words are removed, but what people said is not rewritten.
If the file contains no speech, for example a music track or a recording of room noise, the output says so and describes what it hears instead of inventing words. Long files are sent in a streaming connection that stays open while the model works, so a recording that takes several minutes to process doesn't time out halfway.
The summary comes in the language that's spoken, and the transcript is never translated. A long recording can make the model's answer hit its length limit; when that happens the request is retried automatically, so you don't get a transcript that stops in the middle of a sentence without warning.
MP3, WAV, M4A and OGG files up to 25 MB are accepted, and audio conversion is part of Premium, together with video and YouTube: up to 10 of those conversions and 120 minutes of media a day. The file goes to the model for the length of the request and is deleted when it ends. Nothing is stored after the Markdown is delivered.
How it works
- 1Drop your audio file (MP3, WAV, M4A, OGG — up to 25 MB) in the converter above.
- 2The AI transcribes the speech and structures it: summary, sections, speaker labels.
- 3Preview the Markdown and download the .md file.
Before and after
[14-minute recording] Three voices, some crosstalk. Topics: release status, a bug in production, next sprint.
## Summary Daily standup covering the v2 release, a production bug and sprint planning. ## Transcript **Speaker 1:** The release is on track for Thursday. **Speaker 2:** One blocker — the export bug hit two customers overnight...
Frequently asked questions
Which audio formats are supported?
MP3, WAV, M4A and OGG, up to 25 MB per file.
Does it detect different speakers?
When voices are clearly distinguishable, turns are labeled as Speaker 1, Speaker 2 — or with real names if they're mentioned in the recording.
Is my recording stored?
No. The file is processed only for the duration of the conversion and deleted immediately afterwards — meetings and interviews often contain sensitive content, and we treat them that way.
Is audio to Markdown free?
No. Transcription is done by AI and costs by the minute, so it's part of Premium: files up to 25 MB, up to 10 audio, video or YouTube conversions and 120 minutes of media a day. Converting documents and images stays free.