Result
Your transcript
Listen to the recording and check the words. The file plays from your device.
Download recording as WAVYour browser could not play this recording. Try WAV, MP3 or MP4 with an AAC track.
Choose a file, use the sample or record your voice. Click Transcribe; your editable text will appear here.
SRT subtitles
The model estimates segment timing. Check it against the player. For SRT, edit text in the segment editor; changes also update TXT. Editing the full transcript above disables SRT so that outdated subtitles cannot be exported.
Review segment text and timing
Time is relative to the start of the file, in HH:MM:SS,mmm. The end must follow the start; segments must not overlap. Fill any missing timestamp using the recording. An empty text field excludes the segment from TXT and SRT.
Check names, numbers and details. Selecting another file, sample or recording clears the previous result. Save your recording and transcript before closing this tab.
How to transcribe your recording
- Choose a file containing English speech, or use the sample recording. Audio files can be up to 20 minutes and 50 MB. WAV, MP3 and M4A work when your browser supports their codec.
- Click Transcribe. The first run downloads about 110 MB of recognizer files from Hugging Face. Keep the tab open while the model runs on your device. A computer is recommended.
- Listen to uncertain words in the player and correct the transcript. Check names, numbers and technical terms. Noise, music and overlapping voices can reduce accuracy.
- Copy the edited text or download TXT. To export subtitles, review the segments and download SRT.
Audio, video and microphone recordings
MP4 with AAC audio and WebM with Opus are recommended for video. The container alone does not guarantee compatibility: your browser must support the audio codec, and the video must contain audio. The same 20-minute and 50-MB limits apply. The recognizer reads spoken words, not text in the picture.
To record a voice note, click Record voice and allow microphone access. Click Stop recording when finished. You can listen, download WAV and then click Transcribe. Microphone recordings are limited to 5 minutes and 50 MB. Reaching the limit saves the received audio; an interrupted recording is labeled for review. This records the microphone, not direct system audio.
Files are selected from your device. This tool does not download recordings from YouTube, social platforms or other links. Only process recordings you have the right to use.
Review and download SRT subtitles
The same recognition result supplies the transcript and estimated segment timestamps. Open Review segment text and timing, listen in the player and correct each segment. These are segment estimates, not word-level alignment. The tool does not identify speakers or translate speech.
Use HH:MM:SS,mmm for start and end. Fill missing timestamps by checking the recording. Each non-empty segment must end after it starts, stay within the file duration and not overlap the previous segment. Empty text excludes a segment from both TXT and SRT.
Edits in the segment editor update both TXT and SRT. Editing the full transcript disables SRT so that stale timing cannot be exported. To create timed subtitles again, transcribe the recording again. Selecting another source clears the previous result. Save your files before closing the tab.
SRT requires one audio track with a codec and timeline that your browser can decode. If only TXT is available, the result explains why. Multiple tracks, an unsupported decoder or a browser that removes gaps from the track can prevent reliable timing. Try WAV with pauses preserved or another browser. Very large decoded audio is also bounded; shorten high-rate recordings or export them as 16 kHz WAV.
Download SRT and import it into a video editor or player to check synchronization. This creates a separate subtitle file. It does not burn captions into video or export a captioned video.
Free, on-device processing
Recording, WAV, transcription, editing, copying, TXT and SRT are free in this web tool. There is no account, paid unlock or purchase of minutes. Cuevy's paid desktop call assistant is a separate product; its other pages are currently in Russian.
The browser runs the English-only Whisper base.en model with openly licensed components. Audio, video and transcripts are not uploaded to Cuevy. Usage analytics record page opens, completion and export without file names, recording content or transcript text. Session replay is disabled on this tool.
Questions before you start
- Why does the first run download about 110 MB?
- These are recognizer files, not your recording. They may stay in the browser cache. The English and Russian versions use separate models; using the other language downloads its recognizer on the first run. Clearing the cache or using another browser may also require downloading again. Progress reports the current file, not completion of the entire transcription.
- Does this page recognize Russian speech?
- This page is configured for English speech. For Russian, use the Russian version. Changing pages starts a separate session; save your current recording and result first.
- Can I use a phone?
- You can try a current browser, but recognition uses memory and processing power. A computer is recommended. Start with a short recording and keep the tab open. If the browser interrupts processing, shorten the file or use a computer.
- When does the microphone turn on?
- Only after you click Record voice and grant permission. Stop recording, Cancel request, selecting another source or leaving the page stops microphone use. If a permission request was cancelled, granting it later does not automatically start recording.
- Is the result ready to publish?
- Review it first. Recognition can mishear words, and model timestamps are estimates. Check the recording, names, numbers and synchronization. The tool does not summarize or verify what was said.
Useful to someone else?
Share a link to the free tool. Your answers, files and result are not included.