Skip to content
Marvin's Toolbox.

Audio and Video Transcriber

Turn speech in an audio or video file into text with Whisper. Pick a model size and a language, or translate to English, then read a timestamped transcript and download it as text, SRT or VTT subtitles.

Runs entirely in your browser. Nothing you put in here gets uploaded.

Your file

Drop an audio or video file here

MP3, M4A, WAV, MP4, MOV, MKV and more. Click to choose a file.

It never leaves your browser.

Transcription

Model
Spoken language
Writes English, whatever language is spoken.

Next steps

Send this tool's output straight into another tool.

Subtitle Converter

Convert subtitles between SRT and VTT, shift their timing forward or backward, and fix frame rate drift to sync out-of-sync captions. Preview the parsed cues, catch format errors, and download the converted file.

textconvertervideomedia

Word & Character Counter

Count words, characters, sentences, paragraphs, lines and more as you type, with estimated reading and speaking time.

text

Text Case Converter

Convert text into every case at once: title case in the APA, AP, Chicago, MLA, Bluebook, AMA and NYT styles, sentence case, upper, lower, camelCase, snake_case, kebab-case and more.

textconverterdev

Dedupe Lines

Remove duplicate lines from a list or text, keeping the first occurrence. Optionally ignore case and whitespace, or drop blank lines too.

textdev

Find and Replace

Replace text in two modes: plain find and replace all, or regular expressions with flag controls and capture group references like $1.

textdev

Text Diff

Compare two texts and see every added, removed and changed line highlighted, side by side or inline, with the changed words marked within each line. Ignore case, whitespace or blank lines, read a summary of how much changed, and export the result as a unified diff.

textdev

UUID Decoder

Paste a UUID to see what it carries: version, variant, and for time-based versions the exact timestamp, clock sequence and node. Also shows the raw bytes, the 128-bit integer and the URN form.

devuuid

Base64 Encoder & Decoder

Decode Base64 to text or encode text to Base64. Paste into the top box, read the result below, and flip the direction with one click. Handles Unicode correctly, reads URL-safe Base64, and shows binary payloads as a hex dump you can download.

devconvertertext

Number Base Converter

Convert numbers between binary, octal, decimal, hexadecimal and any base up to 36. Type into any field and the others update as you go. Handles huge integers without losing precision and accepts 0x, 0b and 0o prefixes.

devconvertercalculator

About the Audio and Video Transcriber Tool

This tool turns the speech in an audio or video file into text with OpenAI's Whisper model. Drop in a voice memo, an interview, a lecture or a screen recording, pick a model size, and read a timestamped transcript as it is written. Click any segment to jump the player to that moment and check what was said.

Three model sizes trade speed for accuracy. A model downloads once from Hugging Face the first time you use it, with its size stated before anything starts, and your browser keeps it for next time. Whisper runs on the graphics card through WebGPU where that is available and on the processor through WebAssembly where it is not, and the page shows which one did the work.

What you can do

  • Transcribe an MP3, M4A or WAV recording to text
  • Get a transcript of an MP4, MOV or MKV video
  • Generate SRT or VTT subtitles for a video
  • Translate speech in another language into English text
  • Detect the spoken language automatically or set one of 99 by hand
  • Jump to any line of the transcript in the player
  • Copy or download the transcript as plain text with or without timestamps

How to use the Audio and Video Transcriber

  1. 1Drop an audio or video file onto the page.
  2. 2Pick Tiny, Base or Small. Each one states its download size, and a saved model needs no download.
  3. 3Leave the language on automatic detection or pick it, and switch on translation if you want English text.
  4. 4Start the transcription and watch the segments appear. Cancel at any time and keep what is done so far.
  5. 5Click a segment to check it in the player, then copy or download the text, SRT or VTT.

Which model to pick

Tiny is the quickest and works for clear speech in a quiet room, but it misspells names and gets punctuation wrong more often. Base is a good default for most recordings. Small makes noticeably fewer mistakes, especially outside English, at the cost of a bigger download and a slower run. Models are larger on WebGPU than on WebAssembly because the two engines use different weight formats, so the sizes shown are the ones for your machine. A saved model can be removed again from the same panel to free the disk space.

Long files and silence

Whisper reads 30 seconds at a time, so a long file is worked through window by window, and each window resumes at the last finished sentence so no words are cut at the seams. Progress is measured in audio time, and the text so far appears while the rest is still running. Stretches of silence are skipped without running the model at all, which makes recordings with long pauses faster. Whisper likes to invent a polite "Thank you" over silence and noise, so text the model was unsure of over audio it rated as speechless is dropped, and a file with no speech reports exactly that instead of made up words.

Files the browser cannot decode

Most audio and video formats decode directly and instantly. For containers and codecs the browser has no decoder for, such as AC3 audio in an MKV, FFmpeg extracts the audio instead. If a video has no audio track, the tool says so right away. Very large videos are quicker to handle if you first pull the sound out with the Video Audio Extractor.

Subtitles and what comes next

The SRT and VTT exports split the transcript into cues of at most two lines of 42 characters, the usual limit for readable subtitles. If the timing needs adjusting, or the file needs another format, send it on to the Subtitle Converter. To check the length of the text, the Word & Character Counter takes the transcript as it is.

Credits

Open source does the heavy lifting in this tool. Thank you to:

  • Transformers.js

    Hugging Face's JavaScript library for running machine learning models with ONNX Runtime, on WebGPU or WebAssembly.

  • OpenAI Whisper

    An open speech recognition model trained on 680,000 hours of multilingual audio.

  • ffmpeg.wasm

    A WebAssembly build of FFmpeg, the audio and video toolkit behind half the media software you know.

Everything this site builds on is listed on the credits page.

Related tools

Tools that share a tag with this one.

Subtitle Converter

Convert subtitles between SRT and VTT, shift their timing forward or backward, and fix frame rate drift to sync out-of-sync captions. Preview the parsed cues, catch format errors, and download the converted file.

textconvertervideomedia

Video Audio Extractor

Pull the audio track out of a video as MP3, M4A, WAV, FLAC or OGG. Drop in one or more files, pick a track and bitrate, or keep the original audio untouched when the format allows it, and download the results.

mediavideoaudioconverter

Audio Converter

Convert audio files between MP3, WAV, OGG, FLAC and M4A. Drop in as many files as you like, pick the target format and bitrate, and download the results one by one or all at once.

audiomediaconverter

Audio Inspector

See everything about an audio file: duration, format, codec, sample rate, bit depth, channels and bitrate, the embedded tags and cover art, plus measured peak and RMS levels and a waveform overview.

mediaaudio

Audio Metadata Editor

View and edit the ID3 tags of an MP3, including title, artist, album, year, genre, track number, lyrics and cover art. Reads the existing metadata of other audio formats too, and saves a tagged copy of your MP3.

audiomedia

Audio Trimmer

Cut a section out of an audio file on a zoomable waveform timeline. Drag the start and end handles, play back just your selection, fade the edges if you like, and export the result as MP3, WAV, OGG, FLAC or M4A.

audiomedia