Audio and video to text
Transcribe short audio or video locally, edit captions and save TXT or SRT files.
Audio and video transcription
Select a file or use the public editing example.
Listen to the source
Files and captions stay in this page’s memory and are cleared on refresh or departure. Only model assets may be cached.
Review and edit captions
Recognition can invent or mishear words. Listen to the source and check every caption and timestamp. Speaker identification and translation are not provided.
Instructions and example
Transcribe MP3, WAV, MP4 and WEBM speech in Korean, English or Japanese, then edit the result. Whisper base provides estimates, not guaranteed accurate meeting minutes or professional transcripts.
The public editing example contains “Hello, welcome.” from 0–2 seconds and “Review this caption.” from 2–5 seconds. SRT records 00:00:00,000 → 00:00:02,000 and 00:00:02,000 → 00:00:05,000. Select a real recording to recognize speech and edit its estimated segments in the same way.
Frequently asked questions
Which files and durations are supported?
Files are limited to 60 seconds and 20MB, and the browser must decode their audio codec. Use the audio extraction tool for unsupported videos. Split longer recordings with audio trimming and process one file at a time.
Can the model invent incorrect words?
Noise, music, quiet speech and silence can cause omissions or hallucinations. Check every caption and start/end time against the source. Speaker identification, live recording and automatic translation are not supported.
Is my recording uploaded?
Input files and captions are not uploaded. On your explicit action, model and runtime files download from this site and processing runs in your browser. Refresh or departure clears inputs and results; model files may remain cached.
Calculation criteria and reference materials
- OpenAI — Whisper model card
Describes language performance differences, hallucinations and recognition limits of Whisper. This tool runs the base model in the browser.
- Hugging Face — Automatic speech recognition
API reference for browser speech recognition and estimated caption timestamps. It does not guarantee result accuracy.
Verification of calculation criteria: · Calculation·verification principles · Report errors