Help
From your recording to text you can keep.
How do I import a recording?
Choose an audio file saved on your device, or drop it into the tool. Check its name and duration, select the spoken language, then press Transcribe.
On iPhone, use Share in Voice Memos to save the recording to Files or transfer it to your computer. On Android, use Share or Export in your recorder app. Choose the actual audio file, not a cloud sharing link or a recorder project.
If a recording is only a cloud placeholder, download it first. A laptop or desktop is recommended for longer recordings and larger models.
Which files can I use?
Use MP3, M4A or supported PCM WAV. Compressed audio is limited to 60 minutes because it is decoded in browser memory. Unknown-duration compressed files larger than 100 MB are rejected before decoding.
YouTube links
Choose YouTube link, paste a public video URL and select the spoken language. Get transcript uses captions in that language when available. Otherwise it retrieves the audio and transcribes it on your device. Settings lets you choose audio recognition directly and adjust quality.
Videos must be non-live, available without signing in, no longer than 60 minutes and within the 100 MB audio limit. Private, age-restricted and region-blocked videos are not supported. YouTube can sometimes refuse retrieval; retry or choose another video. An unreadable caption track is reported as an error. Choose the audio route to try recognition instead.
For caption results, timestamps open the corresponding point on YouTube. For audio results, the built-in player lets you listen without leaving the workbench. Use recordings you own or have permission to process.
The hosted YouTube service handles one retrieval at a time, with up to 10 requests per connection per UTC hour and 100 shared requests per UTC day. A busy or capacity message means you can retry later or use an audio file. These limits do not apply to local-file transcription.
MP3
A supported MP3 can be selected directly. Split recordings longer than 60 minutes into shorter files before starting.
M4A
M4A support depends on the codec inside the file and your browser. If decoding fails, export AAC M4A or PCM WAV from your recorder or audio editor. Changing a file extension does not convert the audio.
WAV and damaged files
Supported PCM WAV is decoded in sections. Device memory, speed and encoding still set practical limits. If a file fails, check that it plays in your recorder, export it again, or try a shorter recording on a computer.
Why does a model download? Where does my audio go?
Speech recognition runs on your device. Selected files and your transcript stay in the current page; the original file is unchanged. The browser downloads model and runtime files, and the site sends anonymous usage events without audio, filenames or transcript text.
Fast and Balanced are intended for English. Selecting another language switches to Best. Open Settings before starting to review quality and approximate model download sizes. Larger models need more memory; preparation takes time even after downloading.
The browser may reuse cached models, but caches can be cleared or evicted. Keep the tab open while it works. For download errors, check your connection and retry. For memory problems, try a shorter file on a computer.
The full transcript is free, with no signup or commercial minutes quota. Format and device limits still apply. There is no cloud library or transcript history.
How do I check, edit and save the text?
Search for a phrase and choose Next to move through matching segments. Click a timestamp to listen back. Choose Edit to correct the text, then Done editing. Timestamps cannot be edited.
For lectures, check names and technical terms against the recording. For podcasts, verify wording and timing before using a subtitle export. Speaker labels are not automatically identified. Chapters and extracted key points are optional navigation aids and update when you edit the text.
Copy takes the complete edited text. Export offers TXT for plain text, SRT for subtitles and Markdown for text with timestamps and chapters.
You can cancel during file checks, model preparation or recognition. Completed sections remain a partial transcript that you can edit and export. Retrying starts from the beginning; export anything you want to keep before replacing a result.
Copy or export before refreshing, leaving or closing the page. Files, text and edits only live in page memory and are not saved automatically.