🎙️
Removedor de Ruído de Áudio
Remova ruído de fundo de gravações antes da transcrição
GrátisNavegadorSem Cadastro
🎙️
🌐 Runs in Browser⏱ Est. 1–3× real-timeNEW
Transcrição de Áudio
Transcreva áudio em texto com precisão de nível Whisper — grátis e privado
Try It Now — FreeWhat is Transcrição de Áudio?
Convert speech to text with high accuracy using a browser-based Whisper model. Supports 99 languages, generates timestamped transcripts, and exports to SRT, VTT, or plain text. No upload, no account.
Key Features
Powered by OpenAI Whisper (tiny/base/small models in WebAssembly)
Supports 99 languages with automatic language detection
Word-level timestamps for precise navigation
Export to SRT, VTT, and plain TXT formats
Speaker diarization (beta)
Real-time transcription for microphone input
Edit and correct transcript in-browser
Upload MP3, WAV, M4A, FLAC, OGG, WebM up to 500 MB
How It Works
1
Load Model
On first use, the Whisper model (approx. 150 MB) is downloaded once and cached in your browser.
2
Upload Audio
Drop your audio file or record directly from your microphone.
3
Transcription
The model processes your audio locally in chunks, producing timestamped text segments.
4
Review & Export
Read, edit, and search your transcript, then export in your preferred format.
Who Is This For?
- →Journalists transcribing interviews
- →Podcasters creating show notes and subtitles
- →Students transcribing lectures
- →Content creators generating captions for videos
- →Researchers transcribing qualitative data
Why Use Transcrição de Áudio?
Cloud transcription services like Otter.ai and Descript send your audio to remote servers — a serious privacy risk for confidential meetings, medical consultations, or personal recordings. Zeraku's Audio Transcription runs the entire Whisper model inside your browser using WebAssembly. Your recordings never leave your device. Unlike cloud tools that charge per minute or lock features behind subscriptions, Zeraku is completely free with no account, no usage limits, and no file-size cap beyond 500 MB. Need to transcribe offline? After the first model download (approx. 150 MB) the tool works fully offline from the second visit onwards — reliable even on a train or plane. Typical use cases include: creating meeting minutes from Zoom or Google Meet recordings; generating SRT subtitle files for YouTube uploads; transcribing interviews for journalism or qualitative research; converting lectures and seminars to searchable text. Powered by OpenAI Whisper — the same model behind industry-leading transcription services — supporting 99 languages with automatic language detection.
How Zeraku Compares to Cloud Services
Most transcription tools send your audio to remote servers — a serious privacy risk for confidential meetings, medical consultations, or personal recordings. Zeraku processes everything in your browser so your data never leaves your device.
| Zeraku | Service A | Service B | |
|---|---|---|---|
| Completely free | ✓ | △Up to 3/month | △Up to 10 min |
| Privacy (no data upload) | ✓Browser-only | ✗Server upload | ✗Server upload |
| No account required | ✓ | ✗Required | ✗Required |
| Works offline (2nd visit+) | ✓Cached model | ✗Always online | ✗Always online |
| Supported languages | 99言語Auto-detect | 58言語 | 100言語+Paid plan only |
| SRT/VTT subtitle export | ✓Free | ✓Paid plan only | ✓Paid plan only |
| Speaker diarization | △Beta | ✓Paid plan only | ✓Paid plan only |
| Max file size (free) | 500MB | 25MBFree plan | 100MBFree plan |
Beginner's Guide
Audio transcription converts spoken words into written text automatically. It is useful for turning meeting recordings into minutes, adding subtitles to videos, writing up interviews, or summarising lectures. To use it: upload your audio file (MP3, WAV, M4A, or FLAC), select the spoken language, and press Start. The tool transcribes the audio and lets you download the result as a text file (.txt) or a subtitle file (.srt) that can be uploaded directly to YouTube. No technical knowledge is required.
Technical Details
Powered by whisper.cpp compiled to WebAssembly via Emscripten, running inside a dedicated Web Worker so the UI stays fully responsive during transcription. Audio decoding uses the Web Audio API; input files are normalised to 16 kHz mono PCM before processing regardless of source format. Long recordings are automatically chunked into 30-second sliding windows with 1-second overlap to ensure seamless output across boundaries. The Whisper small model (~150 MB) is fetched once on first use and stored in the browser Cache API, enabling fully offline operation from the second session onwards. Speaker diarisation (beta) applies a lightweight spectral clustering algorithm on audio embeddings to label speaker turns. Output is available as plain text (.txt) or SubRip (.srt) with millisecond-accurate timestamps.
Frequently Asked Questions
Related Tools
Ready to try Transcrição de Áudio?
Open Transcrição de Áudio — Free