Get a readable transcript of a recording, split into paragraphs, ready to copy or save as a text file. Nothing is uploaded, and there’s no account or watermark.
Select the recording: a video (MP4, MOV, WebM, MKV) or an audio file such as a voice memo (MP3, M4A, WAV, OGG).
Pick the “Spoken language” to skip detection, and choose “Fast”, “Balanced” or “Best” depending on how much time you have.
Start with “Generate subtitles”. Lines appear as they are recognised, and “Cancel” stops the job if you picked the wrong file.
Tick “Timestamps in the text file” if you want times before each paragraph, then press “Download TXT”, or “Copy text” to paste it into a document or email.
A transcript is the plain words of a recording, meant to be read rather than shown over a video. filetidy creates it with the Whisper speech model in your browser, then joins the recognised lines into paragraphs, starting a new paragraph wherever the speaker pauses. The result reads like notes rather than like a column of short caption lines.
Audio files are first-class inputs here. A voice memo, a phone call recording exported as M4A, a podcast episode in MP3 or a WAV from a recorder all work the same way as a video, because the speech is extracted and converted to 16 kHz mono before recognition anyway.
With “Timestamps in the text file” ticked, each paragraph starts with the time it begins in the recording, which makes it easy to go back to a quote in an interview or a decision in a meeting. Without it, you get clean text ready for a summary, an article or a study sheet.
Because the file is never uploaded, confidential material such as internal meetings, patient interviews or legal calls stays on your device. The first 10 minutes of each file are transcribed for free.
iPhone & iPad
Transcripts in Voice Memos and Notes need a recent iOS version and support only a few languages, English first; availability varies by region. This works on recordings made on the phone, not on videos.
Android
Recorder is preinstalled on Google Pixel phones; transcription languages are limited. It cannot transcribe a video file you already have.
Mac
English-focused, Apple silicon only and not available in every region. There is no way to save the text as a file.
Subtitles are timed fragments designed to be read in a few seconds each over moving pictures. A transcript is continuous text designed to be read, searched and quoted. If nobody will watch the video with the text on top, you want the transcript.
Download both if you need both. The same recognition produces the SRT for the video and the TXT for your notes, so publishing a captioned lecture and handing out its transcript costs no extra time.
For meetings with many voices, the transcript will not say who is speaking, and overlapping speech is the main source of errors. Recording with a microphone close to the speakers improves the text more than any setting.
Yes. MP3, M4A, WAV and OGG files are accepted alongside video. Voice memos, podcast episodes and dictaphone recordings are transcribed exactly like the soundtrack of a video.
No. The text is split into paragraphs at pauses, which often coincide with a change of speaker, but names or labels are not added. Insert them yourself after pasting the text into a document.
Tick “Timestamps in the text file” before pressing “Download TXT”. Each paragraph then begins with its start time, so you can find the passage in the recording.
The recording is processed by a model running in your browser and is never sent to filetidy or anyone else. Only the speech model and its engine are downloaded, once, and then kept in the browser cache.
The free version transcribes the first 10 minutes. Pro, which is coming soon, is designed for recordings of several hours and is much faster; a waitlist for it will open on the pricing page.
Cross-talk, distant microphones, background noise and jargon are hard for any speech model. Choosing the language and a larger “Accuracy” setting helps; for important quotes, check the passage against the recording.
Last updated October 28, 2026
Choosing the language is faster and more accurate than detection, especially for short clips.
Speech recognition makes mistakes, especially with names, numbers, music and several people talking at once. Dialects such as Moroccan Darija are recognised less reliably than Modern Standard Arabic. Read the subtitles before you publish them.
Need long videos or the best accuracy?
AI subtitles Pro will use Whisper large-v3 on our servers: about an hour of video in a few minutes, better accuracy on accents and dialects, files up to 3 hours. Only compressed audio is sent, never the video.
See Pro and join the waitlist