Pick an AI audio tool by what you actually need

Pick the tool by what you want to end up with, not by the tool's name. The list below matches each need to its tool on TachNhac.com.

Which need calls for which tool?

• ◆You want vocals and backing track apart, or a karaoke beat: Music separation. Choose 2 layers (vocals and backing track) or 4 layers (drums, bass, vocals, everything else).
• ◆You want one section of a track: Audio cutter, which runs in the browser, up to 15 minutes and 100 MB per file.
• ◆A song sits too high or too low for your voice: Song key changer, shifting from 12 semitones down to 12 up (one octave), tempo unchanged.
• ◆You want to practise a hard passage slowly: Audio speed changer, from 50% to 200% of the original speed, pitch unchanged.
• ◆A music file is in the wrong format: Audio converter, exporting MP3, WAV or M4A.
• ◆You suspect a "lossless" file is fake: Spectrogram. A flat black band across the upper half of the plot means the file had its high frequencies cut off.
• ◆You need the audio from a video link: Link downloader.
• ◆You have a sheet music PDF exported from notation software: PDF to MIDI. Photos and scans cannot be read yet.

Listen back after each step

Separation results depend on the source. The separation tool's result screen says plainly that live, layered or heavily compressed recordings can leave other sound behind, and that you can lower the backing track by remixing the separated layers instead of removing it outright. So listen to each result before passing it to the next tool.