AI Transcribe

Transcribe audio to text, with speakers and subtitles

Upload a recording of up to 60 minutes. You get the full text, a label for each speaker, a timestamp on every word and an SRT or VTT file you can drop into your video editor.

The first 10 minutes each day are free. After that you pay 1 credit per started minute.

Speaker 1 · Morning quick check on the new weather app we launch next Thursday,
Speaker 1 · so let's go over the three things we still

Real output from the tool for this 7-second clip. Click a time to jump there. The voice file stops halfway through a sentence, so the text does too.

The subtitle file it made from that clip

Each subtitle stays under 42 characters per line, two lines at most and no longer than five seconds on screen. A new subtitle starts after a full stop or a question mark, or when another speaker takes over. A line that gets too long breaks at the last comma, not halfway through a phrase.

Audio length: 7.01 s
Words with a timestamp: 21 of 21
Speakers found: 1
Cost: 1 free minute

tram-stop.srt

1
00:00:00,180 --> 00:00:04,858
Morning quick check on the new weather app
we launch next Thursday,

2
00:00:04,858 --> 00:00:07,000
so let's go over the three things we still

What does a transcript cost?

We count the length of your file before we start. A 61-second voice memo counts as 2 minutes, because every started minute counts.

File length	First file of the day	Free minutes used up
45 seconds	Free	1 of 10
8 minutes	Free	8 of 10
25 minutes	15 credits	10 of 10
60 minutes	50 credits	10 of 10

If a transcript fails, the credits go back to your balance and the free minutes go back to your daily allowance. Credit prices are on the credit packages page.

Which files work?

Audio and video, up to 60 minutes and 200 MB. From a video the tool only uses the sound.

MP3 · WAV · M4A · AAC · FLAC · OGG · Opus · WebM · MP4 · MOV

You can also paste a public link to a file instead of uploading it. You download the result as plain text, SRT or VTT.

Languages and translation

The tool runs on Nemotron 3.5 Transcribe from NVIDIA. NVIDIA's model card lists 40 languages, and the model picks up the language by itself. On this page we only tested English, so treat the rest as the maker's claim until you try your own file.

The translation menu in the tool has 32 languages: English, Chinese, Japanese, Korean, French, German, Spanish, Portuguese, Italian, Russian, Dutch, Polish, Czech, Slovak, Hindi, Bulgarian, Croatian, Danish, Estonian, Finnish, Hungarian, Latvian, Lithuanian, Romanian, Slovenian, Swedish, Thai, Ukrainian, Arabic, Greek, Hebrew, Maltese.

It can also translate while it transcribes: speech in another language into English, or English speech into another language. With translation on you get the text without word timestamps, so no subtitle file.

FAQ

How do I transcribe audio to text?

Log in, open AI Transcribe, upload your audio or video file or paste a public link, and press start. When it is done you read the text on the page and download it as TXT, SRT or VTT.

Is it free?

The first 10 minutes each day are free on any account. After that a transcript costs 1 credit per started minute.

Can it tell speakers apart?

Yes. With timestamps on, every word gets a speaker label, and a new subtitle starts when another speaker takes over. Our own sample has one speaker, so test a conversation of your own before you rely on it for an interview.

Which languages does it understand?

NVIDIA lists 40 languages for Nemotron 3.5 Transcribe and the model detects the language by itself. We tested English.

Can I translate a recording?

Yes, to or from English. Pick a target language before you start. A translated transcript has no word timestamps, so you get text but no subtitle file.

What is the longest file I can upload?

60 minutes and 200 MB. Cut longer recordings into parts first.

What happens to my file?

If a transcript fails we delete the upload and refund the credits. After a successful run the file stays in your account so the player on the result page works.