| The job | Tool | Guide |
|---|---|---|
| Make a backing track to sing over | Vocal remover | Karaoke tracks |
| Make an interview intelligible | Noise reducer | Podcast cleanup |
| Narrate a short video without recording | Text to speech | Video voiceover |
| Turn interviews, lectures and voice memos into text or subtitles | AI Speech to Text | Transcribe audio |
| Create moving music for headphones | 8D Audio Converter | Create spatial audio |
These tools run in your browser: nothing is uploaded and no account is required. The limits differ by tool — the two audio filters take up to five minutes of audio per pass, and text to speech takes up to 500 characters per generation. What that means in practice, and where it gets in the way, is covered in each guide.
The tools also combine. A noisy voice sample cleaned with the noise reducer makes a better clone than the raw recording. Cleaner audio also improves speech recognition.
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.
Spatial music for headphones
Use 8D Audio Converter to make a song, ambient loop or sound effect move around the listener. Try a slow rotation for music or a fixed angle for a directional effect. Preview with headphones and compare the original mono mix before exporting.
Choose the rotation cycle, direction, starting angle and output volume. Loudness normalization helps compare recordings at a consistent level. Looping affects the preview only; export saves the complete recording once as stereo WAV, MP3 or AAC.
The tool accepts browser-decodable audio up to 100 MB and 10 minutes. Stereo input is mixed into one moving source. The result is a binaural headphone effect, not an eight-channel surround mix. Processing and export stay on your device.
What these uses have in common
Each of them used to mean either buying a plugin or handing the file to somebody else's server. The tools here take a third route: the model is downloaded to your browser and runs on your own processor, which is why there is no upload step, no account and nothing metered.
The limits depend on the task: separation and transcription allow five minutes per file, while 8D Audio accepts ten minutes and 100 MB. These limits keep local processing manageable; no paid tier removes them.
Where these tools are the wrong choice
- Full-length stem separation in one pass. Trimming works for practice and sampling, not for a finished arrangement. A server-side service is the right tool for that.
- Drums, bass and guitar as separate stems. The separator gives vocals and instrumental only.
- Phones and older machines. Local processing is exactly wrong for weak hardware.
- Automation. There is no API. AI Speech to Text supports several files in one batch, with a 5-minute limit per file.
Frequently asked questions
Do Audiofilter’s tools upload my files?
No. Audio is processed in your browser using downloaded code and, where needed, models. Files stay on your device. You can check the network tab while processing.
Why is every tool limited per pass?
Local processing uses your device’s memory and processor. Limits vary by tool: five minutes for separation and transcription, ten minutes and 100 MB for 8D Audio, and 500 characters per text-to-speech generation. There is no paid tier that lifts these limits.