Free AI tool
Transcribe video to text, free, with AI AI
Transcribe video to text with AI.
Your video or audio
Drag a file here, paste it, or choose one.
.mp4, .mov, .webm, .mkv, .m4v, .mp3, .wav, .m4a, .aac, .ogg, .opus, .flac. Up to 500 MB.
Up to 10 minutes long.
Only the audio track is sent, to our server and an AI speech model. It is not stored or used for training.
Pick the language yourself if the video opens with music or silence, or if detection gets it wrong.
Transcript
Worked example
Measured on a 31 second safety briefing, language left on detect.
Audio length
0:31
Transcribed in
4 s
9 cues in all, every word right. Numbers come back as digits ("20kg" for twenty kilograms), so read a transcript through before you publish the subtitles.
1. Drop in a video or audio file. 2. Pick the language. 3. Whisper transcribes the audio. 4. Copy or download.
What Interlace does
Subtitles in 28 languages, from one source.
- Interlace makes the video and its subtitles together
- Voice and on-screen text in 28 languages
- No transcribing and no timing cues by hand
Real Interlace output, unedited. Shown silent and sped up.
Your transcript
00:00:00,000 --> 00:00:02,400
Welcome to the warehouse safety briefing.
Questions
Asked and answered
Only its sound. Your browser takes the audio track out of the file and sends it, as 16 kHz mono audio, to our server, where the Whisper speech model transcribes it on Cloudflare, our hosting provider. The picture never leaves your device. We do not log or keep the audio or the transcript, and it is not used to train AI models.
Most of the wait is the upload: the audio we send is about 1.9 MB a minute, so a 10 minute file is about 19 MB. Reading the audio out of a large video in your browser adds a little time too.
The speech model takes files up to 25 MB, and 10 minutes of audio, in the format we send, is about 19 MB. It also keeps a free tool free. Split a longer recording into parts.
Transcribe it here and press SRT or VTT. Each subtitle holds at most two lines of 42 characters and is timed from the moments Whisper heard the words. YouTube, Vimeo, Moodle and most video players accept both.
Whisper was trained on dozens of languages. The list offers the 15 most asked for, or leave it on detect and the model picks one. If the video opens with music or silence, choose the language yourself.
Turn this into a video
Create an account and start from a document, script, or prompt.