StemSplit Studio 0.2.0

What's new in 0.2.0

Key and tempo, read automatically

Every separation now reports the track's tempo and its musical key, in both
note and Camelot form, so you can beatmatch and mix harmonically without a
second tool. Tempo has ×2 and ÷2 buttons, because which metrical level a
track "is" is genuinely ambiguous — even librosa answers 129 or 64 for the
same track depending on where you start it.

Unusually, this runs on the separated stems rather than the mix: tempo
comes from the isolated drums, key from bass + other with drums and vocals
removed. Chroma is very sensitive to broadband percussion, so having the
stems is a real advantage no tool analysing a finished track can match.

Measured against BPM and key tags written by Serato and rekordbox on a real
DJ library:

	result
Tempo, exact	17 of 18 tagged tracks (94%)
Tempo, within an octave	18 of 18 (100%)
Key, exact	39% — librosa's own CQT chroma scores 35%
Key, exact or harmonically adjacent	68% — librosa scores 56%

Key detection is the weaker half and the app says so rather than pretending:
confidence is shown, and below 0.6 it is labelled a guess, because at that
level it was right only about one time in five.

Play all four stems together

The results pane now plays every stem at once, with mute, solo and a level
fader on each, plus a scrub bar and spacebar transport — so you can check a
mix without leaving the app. Everything stays sample-locked no matter how
much you mute, solo and scrub.

Faster, with bit-identical output

Asking for fewer than four stems is now dramatically quicker on the Maximum
quality setting. htdemucs_ft is four models with one-hot weights — model 0
is fine-tuned for drums, model 1 for bass, and so on — so the app was running
models whose entire contribution it then multiplied by zero and discarded.

job	before	after
Vocals only, processor	52.6s	9.2s
Drums + bass, processor	52.6s	20.1s
Vocals only, NVIDIA GPU	—	1.6s

This is not a quality trade: the output is byte-for-byte identical, because
the skipped terms were multiplied by 0.0 anyway. Asking for all four stems is
unchanged.

Smaller things

• A file passed on the command line now opens, so Open with → StemSplit
  Studio works from Explorer.
• The track's length now shows on the file card before you separate. A
  partial download looks completely normal until the stems come out short.
• Your own logo is in the title bar.
• A file that decodes to no audio now says so clearly instead of failing
  obscurely.

Not changed, deliberately

Demucs' shifts setting was measured against MUSDB18's real ground-truth
stems and left switched off: it buys +0.10 dB mean SI-SDR for 40% more
time, which is far below the ~1 dB where a difference becomes audible.

---

First run: Windows may say "Windows protected your PC" because the app is
not code-signed. Click More info, then Run anyway.

StemSplit Studio 0.1.1

Fixes two problems in 0.1.0.

• M4A/AAC was listed as a supported format but the audio library cannot decode it, so picking an M4A file failed. It is no longer offered, and if one is passed in anyway the app now says to convert it to MP3, WAV or FLAC.
• When the engine reported a problem, a second generic message overwrote it, so failures showed only 'stopped unexpectedly (code 1)'. Real error messages now reach the user.

Supported input: MP3, WAV, FLAC, OGG, AIFF.
Windows 10 and 11, 64-bit.

StemSplit Studio 0.1.0

First release.

Split any song into vocals, drums, bass and other, entirely on your own computer. Nothing is uploaded and no internet connection is needed.

• Demucs v4 (htdemucs) running locally
• Uses an NVIDIA graphics card automatically, falls back to the processor
• 24-bit WAV, FLAC or MP3 output
• Reads MP3, WAV, FLAC, OGG, M4A and AIFF

Windows 10 and 11, 64-bit. Windows may warn that the app is not code-signed - choose More info, then Run anyway.