Transcribe audio and video files to text locally. Export SRT subtitles or plain text. 100% browser-based AI — no upload.
Audio transcription uses OpenAI Whisper models (tiny ~144MB or small ~278MB, fp32) from Xenova, running locally via Transformers.js / ONNX Runtime Web in your browser. Multi-CDN fallback ensures reliable downloads worldwide. Optional GTCRN speech enhancement (WASM SIMD) pre-processes audio to reduce background noise. All models download on first use and are cached afterward. All processing is 100% local — your audio never leaves your device.
This tool runs OpenAI's Whisper model in your browser to turn speech into text. If you enable noise reduction, the audio is first passed through the GTCRN speech-enhancement network (WASM SIMD) to suppress background noise. The cleaned audio is then decoded into mel spectrogram frames and processed by Whisper in ~30-second chunks with a 5-second stride, so each chunk overlaps the last and no words are dropped at boundaries. Whisper's encoder turns each chunk into a representation of the spoken audio, and its decoder emits the text tokens plus start and end timestamps. The output chunks are normalized into readable sentences and can be exported as plain text, CSV, or SRT subtitles. Your recording, the denoiser, and the model all stay on your device.
Audio is optionally denoised with GTCRN, split into overlapping 30-second windows, transcribed by the local Whisper encoder–decoder, and exported as text or SRT.
Priya records a 12-minute interview with a guest in a noisy café. She uploads the M4A, leaves noise reduction on, picks the Small model for accuracy, and starts transcription.
Transcribe audio and video files to text locally using Whisper AI. Everything runs locally; nothing is uploaded.
FreeToolHub Whisper Audio Transcription is a free browser-based tool — transcribe audio and video files to text locally using Whisper AI. No signup, no upload; everything runs locally in your browser.
Transcribe audio and video files to text locally using Whisper AI. Export SRT subtitles or plain text. No upload, no API, free, private.
The Whisper Audio Transcription tool is a free, browser-based AI tool that converts audio and video files to text. It uses OpenAI's Whisper model running locally via Transformers.js with WebGPU acceleration. Upload interviews, podcasts, meeting recordings, or voice notes — the tool transcribes them entirely in your browser. Export as SRT subtitles for video, or plain text for notes. No cloud upload, no API key, no subscription. Your recordings never leave your device.
Journalists transcribing interviews without uploading to cloud services. Podcasters generating show notes and subtitles. Students recording lectures for searchable text. Researchers transcribing focus groups and oral histories. Business professionals who need meeting transcripts but can't use cloud transcription due to confidentiality. Anyone who wants free, private, local audio transcription.
Upload an audio or video file (MP3, WAV, M4A, FLAC, OGG, WebM, MP4 — up to 500MB). Choose a model: Tiny (~40MB, faster) or Small (~240MB, more accurate). Click Transcribe Audio. The model downloads on first use and runs locally. The transcript appears with timestamps for each segment. Export as SRT subtitles, plain text, or copy to clipboard. All processing happens in your browser.
The tool supports files up to 500MB. For long recordings (over 30 minutes), transcription may take several minutes depending on your hardware. The tool processes audio in 30-second chunks with 5-second overlap for accuracy.
Start with Tiny (~40MB) for quick drafts and testing. Switch to Small (~240MB) for important audio where accuracy matters. The Small model handles accents, technical jargon, and overlapping speech significantly better than Tiny.
On clear audio — podcasts, interviews, voice memos recorded close to the mic — Whisper is highly accurate, including punctuation and capitalization. Accuracy degrades with heavy background noise, overlapping speakers, heavy accents, and specialist jargon. The transcript preserves hesitations and filler words in some tiers; for publication-ready text, a light manual edit pass is still expected.
The tiers trade speed against accuracy: tiny is fastest and fine for clear speech and quick drafts; base and small add robustness against noise and accents; medium handles the hardest audio but takes noticeably longer to download and run. If you transcribe regularly, save your tier preference — the model caches locally after the first download, so subsequent runs start instantly at your chosen size.
Common audio containers (mp3, wav, m4a, ogg, flac) and video files (mp4, webm, mov) all work — the audio track is extracted and decoded locally before transcription. For long recordings, processing time scales with duration; a one-hour file takes several minutes even on fast tiers. The transcript is produced as timestamped chunks you can review and copy.
No. The Whisper model runs entirely in your browser via WebAssembly/WebGPU — audio is decoded and transcribed on your device, and nothing is sent to a server. This makes the tool safe for confidential meetings, medical or legal recordings, and any audio you would not upload to a cloud service. Your model-tier preference is the only thing stored locally.
This tool is also known by these tasks — each link opens the same tool with a focused guide:
Why did the cookie go to the doctor?
No signups, no data sold. Every tool is free to use — the free tier allows 5 downloads or saves per day, and the optional Pro plan ($7/mo, $59/yr) adds unlimited downloads, batch processing, white-label exports and an ad-free experience.
☕Support me on Ko-fi— support the free tier100% of proceeds go towards hosting & building more free tools.