Transcribe Audio and Video to Text with a Local Whisper Model

Local transcription with Whisper, no upload

This tool uses a multilingual Whisper model running entirely in your browser, through WebGPU when available (with an automatic fallback to CPU processing). The audio or video file is never sent to our server: audio extraction, normalization, and inference all happen locally on your device.

Only the AI model files (Whisper weights) are downloaded from Hugging Face the first time you use the tool. After that, the browser caches those files, so future transcriptions do not require a new download.

Audio or video, with automatic audio-track extraction

You can upload audio files (MP3, WAV, M4A, AAC, OGG, OPUS, FLAC) or video files (MP4, MOV, WEBM, MKV, and others). When the file is a video, only the audio track is extracted and processed; video frames are not used for transcription.

The tool first tries to decode audio using native browser APIs. If the format or codec is not directly supported, it falls back to a local converter (FFmpeg compiled to WebAssembly) to extract the audio before transcribing, still without relying on any upload.

• Choose a quality tier (Fast, Balanced, or High quality).
• Pick a language manually or let Whisper auto-detect it.
• Transcribe in the original language or translate to English.
• Track real model-download progress and per-chunk transcription progress.
• Cancel at any time without freezing the page.

Timestamps, synced player, and export

The result includes per-segment timestamps, shown both in the continuous text and in the segment list. Click any timestamp to jump the audio/video player to that exact moment.

You can copy the plain text, copy it with timestamps, or export as TXT, timestamped TXT, SRT (subtitles), WebVTT, or a structured JSON file with segments, language, and duration.

Performance, limits, and privacy

On devices with WebGPU support, transcription runs GPU-accelerated. Without WebGPU, the tool uses CPU processing (WebAssembly), which works in virtually any browser but is slower, especially on phones.

Very long files require more browser memory; there is a practical duration limit to avoid freezing the tab. No audio, video, or transcribed text is stored on our servers or sent to analytics.