An independent developer's project
Audiofilter is built and maintained by an independent developer.
I want more people to enjoy technology. Audiofilter gives you practical audio tools to experiment with in your browser.
All tools and exports are completely free, and you can use them without creating an account.
Contact us
If you have additional needs or feature requests, please email us. I will do my best to help.
For bug reports, questions or corrections, email us with the tool name, your browser and what happened. Please avoid sending private recordings.
Follow our demonstrations on X
What I make
Audiofilter brings free audio tools into your browser for music practice, spoken recordings, video narration, speech transcription and creative audio effects.
- AI Vocal RemoverSeparate a song into vocals and instrumental.
- AI Noise ReducerReduce background noise in speech recordings.
- AI Text to SpeechGenerate English speech with a preset or cloned voice.
- AI Voice ChangerChange the voice in a short recording using a reference sample.
- AI Speech to TextTranscribe recordings in multiple languages, edit the text and export TXT, SRT, VTT or JSON.
- Convert
Audio ConverterConvert audio to MP3, WAV or AAC, or reduce the bitrate to compress a recording.
Audio MixerOverlay selected audio segments, with individual volume and timeline placement.
Audio CutterKeep a selected segment or cut out an unwanted section using waveform handles.
- Audio EffectsLocal audio effects: equalizer, reverb, bass boost, nightcore, slowed mixes, pitch-preserving speed and volume boost.
Transcripts and subtitles from recordings
Use AI Speech to Text to transcribe interviews, lectures, meetings or voice memos. Add audio files or record your microphone, review and edit the transcript, then export TXT, SRT, VTT or JSON.
Each file can be up to 5 minutes long. Several files can be processed in one batch. Models are cached in your browser, and your audio and transcript stay on your device.
Use SenseVoice Small for Chinese, English, Cantonese, Japanese and Korean, and Whisper Base for other languages. Automatic language detection can be inaccurate, especially with noise; choose the language manually when possible and review the result.
For noisy recordings, reduce background noise first.
Why processing stays local
Your browser downloads the engine and models, then processes your audio or text on your device. Your files stay on your machine.
How I build the tools
I build the interface and browser integration around open-source audio models. My work includes local processing, model caching, recording, waveform previews and audio export.
Our engineering notes explain the work behind the site, including model delivery, browser isolation and deployment problems.
Open-source foundations
Spleeter- Vocal and instrumental separation.
sherpa-onnx- Browser inference for the separation models.
DeepFilterNet3- Speech noise reduction.
PocketTTS- Text to speech and voice cloning.
OpenVoice2- Voice conversion.
CrispASR, SenseVoice and Whisper- Local multilingual speech recognition and language detection.
FFmpeg.wasm- Audio conversion and export.
These projects provide the models and engines behind the tools. Their authors retain credit for their work; source links are provided above.
Files, storage and website services
Audio and typed text are processed locally. Models and settings can be cached in your browser for later visits.
The text-to-speech tool saves cloned voices and their audio samples in browser storage. You can remove a saved voice or clear site data.
Hosting, analytics, advertising and optional embedded videos involve separate website requests. Our privacy policy explains these services.
What to expect
Results depend on your recording and your device. Each tool page explains its length limits, browser requirements and output formats.
Text to speech supports English. Translating the interface does not change the model's speech language.