There is no paid tier, no trial that expires, no credit card, no email address to hand over, no watermark on the output and no daily quota. You open the page, drop in a file, and download the result.
Why it is free
Because it costs almost nothing to run. Every other tool in this category uploads your file to a server, runs a model on rented GPUs, and has to charge for that compute. Audiofilter sends the model to you instead and runs it in your browser, on your own processor. What that leaves us paying for is bandwidth to deliver a model file once, which a CDN does cheaply.
The architecture that makes it private is the same one that makes it free. They are not two features; they are the same decision.
What you get
| Tool | Limits | Output |
|---|---|---|
| AI Vocal Remover | Up to 5 minutes per track, under 300 MB. MP3, WAV, FLAC, M4A, AAC, OGG and Opus in. | Isolated vocals and a karaoke-ready instrumental, as WAV, MP3 or AAC. |
| AI Noise Reducer | Up to 5 minutes per recording. Same input formats. | The cleaned recording, as WAV, MP3 or AAC. |
| AI Text to Speech | Up to 500 characters per generation. 19 voices, or clone one from a short sample. | Generated speech as a downloadable audio file. |
| MP3 Compressor | Up to 300 MB. MP3, WAV, FLAC, M4A, ALAC, AAC, OGG, Opus, WMA, AIFF, AC3 and CAF in. | MP3 at 128 to 320 kbps, 16-bit WAV, or AAC. |
| AI Voice Changer | Up to 10 seconds of source audio and 30 seconds of reference voice. | The source recording re-voiced as the reference speaker, as WAV, MP3 or AAC. |
| AI Speech to Text | Up to 5 minutes per file. Batch transcription, microphone recording and manual language selection. | Editable transcripts and TXT, SRT, VTT or JSON exports. |
The honest limits
The caps above are real constraints, not upsell hooks — there is nothing to upgrade to:
- Five minutes. Separation runs on your CPU. A full album side would take long enough that most people would assume the page had hung. If you need a longer track, trim the section you want first in any editor and process that.
- Ten seconds, for the voice changer. It needs about 18 seconds of computing for every second of audio, so ten seconds in is already about three minutes of waiting. That, not a paid tier, is where its cap comes from.
- The first load is large. The model has to reach your browser before anything can run, once — after that it is cached and works offline. The sizes differ a lot: about 102 MB for the voice changer, 64 MB for text to speech, 52 MB for the vocal remover, 16 MB for the noise reducer and 2 MB for the converter.
- AI Speech to Text uses Whisper Tiny for language detection and SenseVoice Small or Whisper Base for transcription. Only the needed models are downloaded and cached for later visits.
- Your device matters. A recent laptop handles a
two-minute track in well under a minute, and the time scales with the
length of the audio. An older or memory-constrained
device will be slower, and the text-to-speech tool needs a browser that
supports
SharedArrayBuffer. - Results are not perfect. No separation model is. Dense mixes and heavily processed vocals leave artefacts.
Frequently asked questions
Is it really free, or free for now?
Free, and there is no paid plan behind it to graduate to. If that ever changes, the change will be visible on this page rather than arriving as a locked feature.
Do I need an account?
No. There is no signup, no login and no email capture anywhere on the site. See Privacy for what that means for your data, which is: there is almost none of it.
Is my audio uploaded?
No. All six tools run in your browser tab, on your own device. You can confirm it in ten seconds — open your browser's network tab before you process a file and watch that nothing is sent.
Is there a watermark on the output?
No. The files you download are the unmodified output of the model.
Can I use the output commercially?
We place no restriction on your results and claim no rights over them. Whether you may use a given result depends on the rights in the source material, which the tool does not change — see Terms, section 2.
Is there an API or a batch mode?
No. The tools are browser-only by design, one file at a time. For batch processing or an API, a server-based service is the right tool.