Audio Transcriber
Audio or video file to text and subtitles: no account, no language to pick
Click or drop a media file hereChoose a media file (audio or video: MP3, M4A, WAV, MP4, MOV…)
Max 25 MB sent per file: larger files and other formats are converted on your device before upload
About Audio Transcriber
Audio Transcriber turns the speech in a recorded audio or video file into text with OpenAI's Whisper model. Pick a file and transcription starts at once; Whisper works out the spoken language, so there is nothing to set. FLAC, M4A, MP3, MP4, MPEG, MPGA, OGG, WAV, and WEBM files up to 25 MB are sent to our server as they are. Any other format (AAC, WMA, AIFF, AMR, OPUS, MOV, MKV, AVI…) and any larger file is first turned on your device into an MP3 of its sound that fits under 25 MB, and only that MP3 is sent to our server. A file of up to 4 MB is sent straight through our server; a larger one is first uploaded in parts to our media service. You get the transcript as read-only text, plus TXT, SRT, and VTT downloads with start and end times. It does not record from a microphone; for live dictation, use Audio to Text.
Example
The SRT and VTT files for a voice memo named interview.m4a, built from two timed segments as our server returns them (illustrative text). Downloads are named after your file: interview.txt, interview.srt, interview.vtt, or interview-transcript.zip with all three. When an iPhone has named the file with a code (81414891159__EE8F0326-….MOV, for example), the downloads are named after the date and time of the transcription instead: transcription-2026-10-19-1430.txt, .srt, .vtt, and .zip.
Timed segments (start and end in seconds)
{ start: 0, end: 4.2, text: " Thanks for joining us today." }
{ start: 4.2, end: 9.86, text: " Let's start with how the project began." }interview.srt and interview.vtt
1 00:00:00,000 --> 00:00:04,200 Thanks for joining us today. 2 00:00:04,200 --> 00:00:09,860 Let's start with how the project began. --- WEBVTT 00:00:00.000 --> 00:00:04.200 Thanks for joining us today. 00:00:04.200 --> 00:00:09.860 Let's start with how the project began.
How to transcribe an audio file
- 1Click the upload area and choose an audio or video file, such as MP3, M4A, WAV, or MP4.
- 2Wait while "Transcribing…" is shown, after "Preparing the audio on your device before it is sent" for a file the page converts first; there is no button to press.
- 3Read the text under "Transcript" and click "Copy" to copy it.
- 4Click "Download" next to the .txt, .srt, or .vtt file, or "Download all" for a ZIP of the files.
Formats and limits
- Input
- Audio and video files: OpenAI's formats under the limit are sent to Whisper unchanged; other formats and larger files are sent as an MP3 of their sound made on your device
- Formats
- Sent as they are when under the size limit: FLAC, M4A, MP3, MP4, MPEG, MPGA, OGG, WAV, and WEBM (the list OpenAI publishes). Converted to MP3 on your device before sending, like any larger file: AAC, WMA, AIFF, AMR, OPUS, AC3, CAF, MKA, and the other audio formats the page's ffmpeg engine reads, and the sound of videos such as MOV, MKV, AVI, WMV, FLV, 3GP, and MPG
- Maximum file size
- 25 MB sent per file (about 108 minutes of sound once converted); a file converted on your device before it is sent can be up to 750 MB on a computer and 300 MB on a phone or tablet
- Output
- Text, TXT, SRT, and VTT (subtitles when Whisper returns timed segments)
- Language
- Detected by Whisper; the page has no language setting
- Speaker names
- Not added: all voices in one running text
- Account
- None: no sign-up and no email. Choose, drop, or paste a file and transcription starts
- Usage limits
- Transcriptions per connection are limited per hour and per day, in one allowance with the site's other tools that rely on a paid service, under a monthly site budget; large files also count toward the upload service's own limits
Common uses
- An interview to quote
- A phone interview saved as an M4A voice memo becomes text with the exact wording to quote, and the TXT file goes straight into the draft of an article.
- Captions for a talk or a podcast video
- Choose the video of the talk itself: an MP4, MPEG, or WEBM under the size limit is sent to our server as it is; any other video, or a larger one, has its sound taken out on your device and only that sound is sent. Import the SRT file into a video editor or upload it to YouTube; fix names in the file before publishing.
- Lecture notes
- An MP3 of a lecture becomes text to search and annotate; for a short version, paste up to 8,000 characters of it into Text Summarizer.
- A recording in another language
- A Spanish or French recording is transcribed in Spanish or French, since Whisper works out the language; paste the text into AI Translator for an English version.
Which tool to use
- Audio to Text: Audio to Text adds a microphone mode that types while you speak, with the speech recognition built into your browser (Firefox has none); its file mode sends a recording to the same Whisper model as this page.
- Subtitle Generator: Subtitle Generator listens to nothing: you type each line with its start and end time and get SRT and VTT files. It suits writing or correcting subtitles by hand, while this page writes them from speech.
- Video to Audio: This page takes a video directly and keeps its sound only for the transcript. Video to Audio saves that sound as a file to keep, as an MP3 or in another of its formats.
Where your audio is processed
The speech is not transcribed in your browser but by OpenAI's Whisper API, through our server. A FLAC, M4A, MP3, MP4, MPEG, MPGA, OGG, WAV, or WEBM file under the limit is sent to our server as it is. Any other file, or a larger one, is first converted in your browser, on your device, and only the MP3 of its sound made there is sent to our server, never the original file or its pictures. A file sent up to 4 MB goes from our server to OpenAI; a larger one is first uploaded in parts to our media service on Railway, read from there by our server, and deleted from that service as soon as the transcription is done, fails, or is canceled; if the page is closed or the connection is lost before then, it is deleted after about 15 minutes. We store neither the audio nor the text.
Frequently asked questions
How large can the audio file be?
25 MB is sent per file, the most OpenAI's Whisper accepts in one request. A larger file, or one in another format, is first converted on your device into an MP3 of its sound that fits under 25 MB, and only that MP3 is sent to our server: recordings of up to about 108 minutes, from files of up to 750 MB on a computer and 300 MB on a phone or tablet. Split a longer recording into parts and transcribe them one by one.
Can I make subtitles for a video?
Yes: choose the video itself. An MP4, MPEG, or WEBM video under the size limit is sent to our server as it is; any other video, or a larger one, has its sound taken out on your device, and only that MP3 is sent to our server. Each line of the SRT and WebVTT files carries the start and end time of one of Whisper's segments, ready for YouTube, VLC, or a video editor.
Do I have to choose the language?
No. The page sends no language setting, and Whisper works out the spoken language itself. The transcript is written in that language and is not translated; for a translation, paste the text into AI Translator.
Is it free, and do I need an account?
It is free and needs no account or email. Each connection gets a set number of requests per hour and per day, shared with the site's other tools that rely on a paid service; when one runs out, the page says how many minutes to wait. All visitors together also share a monthly budget: if it runs out, the tool stops until the first day of the next month (UTC) and the page says so.
Does it accept MP3, WAV, and M4A files?
Yes. Under the size limit, our server hands your recording to Whisper untouched, and when Whisper cannot read a file the page says "The AI service couldn't read this file. Try another file or another format." Encrypted music downloads, from Apple Music or QQ Music for example, are stopped before upload with an explanation.
Which audio and video formats does the page accept?
OpenAI's Whisper takes FLAC, M4A, MP3, MP4, MPEG, MPGA, OGG, WAV, and WEBM: those files are sent to our server as they are when they are under the size limit. Every other format the page's ffmpeg engine reads is converted on your device first, and only the resulting MP3 is sent to our server: AAC, WMA, AIFF, AMR, OPUS, AC3, CAF, MKA, and the sound of videos such as MOV, MKV, AVI, WMV, FLV, 3GP, and MPG. A file the engine cannot read is refused before anything is sent.
How do I transcribe a recording that is over the limit?
A file over 25 MB is converted on your device into a compact MP3 before it is sent to our server, so it only has to be cut when it lasts more than about 108 minutes: a 40-minute interview saved as a CD-quality WAV of about 404 MB is sent as an MP3 of about 18 MB (on a computer; a phone takes files up to 300 MB). Cut a longer recording, a 3-hour lecture for example, into two 90-minute halves with Split Audio and transcribe each one. The times in each part's SRT and VTT files start at 00:00:00, so before joining the subtitles, add 90 minutes to every time of the second file.
How accurate is the transcript?
We publish no accuracy percentage: it depends on the recording. Clear speech close to the microphone comes out best; music under the voice, people talking over each other, phone-line audio, and rare names or jargon cause most errors. Over long silences or music, Whisper can write a line twice or write words nobody said, so check the subtitles at those points before publishing them.
Does OpenAI keep my audio?
We keep neither the audio nor the text. The file sent (yours, or the MP3 of its sound made on your device) goes to OpenAI's transcription API; OpenAI's data controls page lists that endpoint as not used for training and with no abuse-monitoring retention (checked October 19, 2026).
Can I edit the transcript on the page?
No. The transcript box is read-only. Copy the text, or download the TXT file, and correct names and technical terms in any text editor or word processor.
How many minutes of audio fit under the size limit?
For a file sent as it is, it depends on the encoding, not on the length: at 25 MB, an MP3 at 128 kbps holds about 27 minutes and one at 64 kbps about 54 minutes, while a CD-quality WAV (16-bit, 44.1 kHz, stereo) holds only about 2.5 minutes. A larger file is converted on your device before it is sent, to a mono MP3 at 64 kbps (up to about 54 minutes in 25 MB), 48 kbps (about 72 minutes), or 32 kbps (about 108 minutes), so the real limit is the length: about 108 minutes.
Does it label who is speaking?
No. Whisper returns one running text without speaker names, so a two-person interview comes back as a single block; add the names yourself in the TXT file.
Where are the timestamps?
Only in the subtitle files: each SRT or VTT line carries the start and end time of one Whisper segment. The text on the page and the TXT file have no times, and when Whisper sends no segments, only the TXT file is offered.
Tips and tricks
- ✓If a recording lasts more than about 108 minutes, cut it into parts with Split Audio and transcribe them one by one.
- ✓Whisper hears speech at 16 kHz in mono, so a high bitrate adds megabytes, not accuracy: a 128 kbps MP3 takes just under 1 MB per minute, while a CD-quality WAV takes about 10 MB.