Video to Text: Get the Words Out of Any Video
Turn what is said in a video into text you can read, search and paste. Take it as a plain transcript, or timed line by line if you need to find the moment something was said. 99 languages, free, no sign-up.
Your video never leaves your device. How?
In Accurate mode only the audio is sent to our server to be transcribed, then deleted. Choose "On this device" in the settings and nothing is uploaded.
Select or drop a video or audio file
MP4, WebM, MOV, MP3, WAV, M4A, FLAC, OGG. Your video stays on your device.
Click, tap, or drop a fileHow to turn a video into text
Choose the video
Any video or audio file on your device.
Pick the language
Or leave Auto-detect on. Naming the language yourself is more accurate.
Read and correct
The words appear line by line. Fix anything misheard.
Take the text
Download a plain TXT transcript, or an SRT if you want the timings with it.
What the text is made from
Only the sound matters here, so the picture is left alone entirely. The sound is read out by Whisper large-v3-turbo, which handles 99 languages and is noticeably better than a small on-device model on anything other than clear English.
The result keeps the sentence breaks the speech actually had, which is why the transcript reads as paragraphs rather than as one unbroken block of words.
Features
Readable, not a wall
The transcript keeps the breaks in the speech, so it reads as sentences instead of one long line.
With or without times
Plain text to paste into notes, or a timed file when you need to jump back to the moment.
Good at other languages
Whisper large-v3-turbo is clearly stronger than an on-device model for Hindi, Arabic, Spanish and Turkish.
Nothing uploaded
Your video never leaves your device. In Accurate mode only the sound is sent to be written out, and nothing is kept.
Frequently asked questions
Can I transcribe a video to text with timestamps?
Yes. The SRT and VTT files have each line with its start and end time, and the plain text comes as a TXT file too. For a transcript with show notes, use Podcast Transcripts.
How accurate are the AI-generated subtitles?
Accurate mode uses Whisper large-v3-turbo, one of the most accurate open speech recognition models, and is clearly better than the on-device model, especially for languages other than English such as Hindi, Arabic, Spanish or Turkish. Accuracy also depends on audio clarity and background noise. You can edit any mistakes before exporting or burning in.
What languages are supported for subtitle generation?
Whisper supports 99 languages including English, Spanish, French, German, Italian, Portuguese, Japanese, Chinese, Korean, Hindi, Arabic, Russian, Dutch, Turkish, Polish, Swedish, and many more. Use Auto-detect or select the language manually for better accuracy.
Does it work offline?
Accurate mode needs a connection. On this device, after the AI model has downloaded once, subtitle generation works offline. The model is cached in your browser, so subsequent visits do not require an internet connection. Burning subtitles into video also works offline once the video engine (about 9 MB) has downloaded.
More from Auto Subtitles
Related: SRT Generator · Transcribe Audio to Text · Podcast Transcripts · Auto Subtitles · All tools