Are separation and downloads free?
Yes. Both vocal and instrumental MP3 and WAV downloads are free, without an account, credits or an added DojoClip watermark. Both tracks come from one separation.
Make an instrumental backing track or extract the vocals from a mixed song. Test a short passage, then separate the full recording on your device. Free downloads, no upload, account or added watermark.
Processed on your device · No account · No added watermark
Choose a song or video
Some vocals or instruments may remain in the other stem, and processing can change the sound. Listen before downloading.
Downloads about 45 MB for the model, plus processing libraries. Your recording stays on this device.
The model is cached when browser storage allows. Your recording is not saved with it.
For a two-minute recording whose chorus starts at 45 seconds, begin the preview there. Listen to Instrumental for the lead vocal and backing harmonies that remain, then compare the drums and chords with Original. If the result suits rehearsal, separate the full recording to get the complete backing track. A useful practice mix may still have artifacts that are noticeable when heard alone.
Video export preserves the encoded picture and replaces its soundtrack. The selected stem may be lowered in volume to prevent clipping. If the video codec cannot be exported, choose MP3 or WAV to download the completed audio.
Trim audio or trim a video before separation. For hiss around spoken dialogue, use background noise removal; for a completely silent video, remove its audio track. Turn a backing track into a music visualizer video, or browse all free tools.
For instrument practice, use the free stem splitter. Choose audio or video to separate it automatically into vocals, drums, bass and other instruments. Leave the tracks you want checked, adjust their levels or loop a passage, then download the same mix as audio or video.
Yes. Both vocal and instrumental MP3 and WAV downloads are free, without an account, credits or an added DojoClip watermark. Both tracks come from one separation.
No. SCNet runs locally. The first use downloads about 45 MB of model data plus processing libraries. The model may be cached for later use; that cache does not store your recording.
Choose a local file up to three minutes and 100 MiB, with browser-decodable mono or stereo audio at 8–96 kHz. MP3, WAV, FLAC, M4A, MP4 and WebM are examples; codecs determine compatibility. The whole-file limits also apply before a short preview.
Choose MP3 for a smaller download or WAV for editing. Both vocal and instrumental files are stereo at 44.1 kHz. MP3 uses 192 kbps and may lower a stem's volume if needed to avoid clipping; WAV keeps the existing 32-bit float samples unchanged. Switching formats reuses the completed separation. For video input, complete the full separation before exporting a compatible MP4 or WebM with the original picture and the chosen stem; additional audio and subtitle tracks are omitted.
It creates two parts: vocals and the combined instrumental. Use Instrumental for singing practice or Vocals for an acapella-style track. It does not generate lyrics, change key, clone voices, split individual instruments or isolate each singer.
Separation estimates overlapping sounds. Reverb, harmonies and dense mixes can leave bleed or remove useful detail. Compare both tracks with the original; results are not guaranteed to match studio stems.
A current desktop browser is recommended. Local processing uses substantial memory and time, so phone performance is not guaranteed. Start with a preview of up to 15 seconds, keep the tab open and download before leaving.
No. Select a file already on your device. The tool does not fetch streaming media. Trim longer local recordings first to fit the input limits.