# AudioBuff

> Free browser-based audio enhancement tool. EQ, loudness normalization, compressor, fades, silence removal, and Whisper-powered transcription — all processed locally on the user's device. Audio never leaves the browser, no account is required, no usage limits, no watermarks.

AudioBuff is a privacy-first web application built on the Web Audio API. All processing (EQ, loudness, encoding, transcription) runs client-side in the browser. There is no server-side audio handling. Suitable for podcasters, musicians, video creators, and anyone who needs quick audio polish without installing software or uploading files to a third-party service.

## Core capabilities

- **8-band EQ**: peaking and shelf filters with one-click presets (Clarity, Warm, Bright, Podcast Voice) plus custom user presets stored in localStorage.
- **High-pass filter**: removes low-frequency noise (rumble, AC hum, mic handling).
- **Loudness normalization (LUFS)**: targets -14 LUFS (Spotify/YouTube), -16 LUFS (Apple Podcasts), or -23 LUFS (EBU R128 broadcast). RMS-based approximation with soft-knee limiter to prevent clipping.
- **Compressor**: dynamics compression to even out volume.
- **Silence removal**: automatically detects and removes pauses ≥2s while preserving 500ms of natural breathing room.
- **Fades**: configurable fade-in/fade-out (0–5s).
- **Trim**: waveform-based start/end trimming with draggable handles.
- **Export**: MP3 (96–320 kbps via lamejs) or WAV (lossless).
- **Transcription**: in-browser Whisper via Transformers.js. Models: whisper-small (240M, ~244MB) and whisper-large-v3-turbo (809M, ~809MB). WebGPU-accelerated where available, WASM fallback. Supports Japanese and English. Outputs SRT, VTT, or plain text.
- **Transcript → cut workflow**: click any transcribed segment to mark it for removal during export.

## Privacy model

- **No upload**: audio files are never sent to any server.
- **No account**: zero user data collection.
- **Local storage only**: presets and preferences stored in browser localStorage; clearing browser data removes them.
- **Whisper models**: downloaded once from Hugging Face CDN and cached in browser; subsequent runs work offline.

## Supported languages (UI)

- Japanese (ja)
- English (en)

## Key URLs

- [Landing page (English)](https://audiobuff.app/en) — product overview, features, FAQ
- [Landing page (Japanese)](https://audiobuff.app/ja) — same content, JA
- [Web app — audio processing](https://audiobuff.app/en/app) — EQ, loudness, transcription, export
- [Web app — voice generation (Beta)](https://audiobuff.app/en/app/voice) — Chatterbox-based browser voice cloning + TTS
- [Web app — transcription](https://audiobuff.app/en/app/transcribe) — Whisper transcription (redirects to /app)
- [Guide index](https://audiobuff.app/en/guide) — all how-to articles
- [How to use](https://audiobuff.app/en/guide/how-to) — step-by-step basic walkthrough
- [EQ & audio processing guide](https://audiobuff.app/en/guide/eq) — EQ theory, filter types, frequency bands, compressor, fades
- [Output settings guide](https://audiobuff.app/en/guide/output) — LUFS targets, MP3 vs WAV, bitrate selection
- [Transcription guide](https://audiobuff.app/en/guide/transcribe) — Whisper internals, model selection, WebGPU acceleration
- [Subtitle formats guide](https://audiobuff.app/en/guide/transcribe-formats) — SRT vs VTT vs TXT, transcript-to-cut workflow
- [Voice generation guide](https://audiobuff.app/en/guide/voice) — Chatterbox internals, reference tips, expressiveness, watermark & privacy
- [Changelog](https://audiobuff.app/en/changelog) — release notes
- [Privacy & terms](https://audiobuff.app/en/legal) — privacy policy and terms of service

## Technical stack

- Next.js 16 (App Router, RSC) on Vercel
- React 19 + TypeScript + Tailwind CSS 4
- Web Audio API (`OfflineAudioContext`, `BiquadFilterNode`)
- lamejs for MP3 encoding (Web Worker)
- @xenova/transformers (Transformers.js) for Whisper
- All audio DSP runs in Web Workers to keep the UI thread responsive

## Pricing

- Free, no account required, no caps, no watermarks.
- Commercial use is welcome.

## Browser requirements

- Modern browser with Web Audio API and Web Worker support.
- WebGPU-capable browser recommended for fast transcription (Chrome/Edge/Safari Tech Preview); falls back to WASM otherwise.
- Mobile (iOS Safari, Android Chrome) is supported.

## Maintainer

- Hooligan Inc.
- Contact: via the in-app contact form at /contact (or the locale-prefixed equivalent).