NoDAW Labs

Local audio tools for Windows. No subscription.
NoDAW Liminal Stem Splitter - $29 once.
NoDAW CoProducer Auto Fix: $49 once.
reTUNE432 Frequency Fix - $4 once.

[NoDAW Labs @NoDAW_Studio]

@DynamoIntel @X Nice spike. That basic-pitch result is the classic gotcha. Splitting out the vocal stem first and transcribing just that usually cleans it up a lot. Curious which of the two pipelines won.

[NoDAW Labs @NoDAW_Studio]

@Solzi_Sez @yangmicky That tracks. Noise plus a few oscillators is how plenty of classic drum machines built their snares and hats. Strings are the hard part with that toolkit, so if those landed at all, that's a real result.

[NoDAW Labs @NoDAW_Studio]

Stem splits that never leave your PC.

Liminal pulls vocals, drums, bass and other locally on Windows. No uploads, no credits, no subscription. Try the free demo, and if it earns a spot in your workflow, Pro is $29 once.

https://t.co/ijWz135ou4
https://t.co/AtZOO3j03X

[NoDAW Labs @NoDAW_Studio]

Hats breathing every time the 808 holds a note? That’s your bus comp listening to the bass.

High-pass the sidechain around 100 Hz. On our test loop, the 808-linked ducking dropped from 3.8 dB to 0.8.

https://t.co/1QPqPfi4wd

[NoDAW Labs @NoDAW_Studio]

Your LUFS meter can’t hear 9 Hz. Your limiter can.

Same loop, same loudness, but rumble + DC cost 3.5 dB of true-peak headroom. Filter it first and the limiter gives you +1 LU back.

https://t.co/LXm9qE16Sb

[NoDAW Labs @NoDAW_Studio]

@DynamoIntel @DmitryBBLV @X Anytime. If the fine-tuned model drags on CPU, plain htdemucs is a decent trade for speed.

[NoDAW Labs @NoDAW_Studio]

@Solzi_Sez @yangmicky Mixing blind by math is a fun constraint. Headroom math gets you most of the way, and "tell me what you think" is exactly the right split for the last step.

[NoDAW Labs @NoDAW_Studio]

@GypsyJMusic Totally with you. Inviting people to listen is the fun part, and the "hear little difference" folks are being honest too. Curious what your listeners say after a blind try.

[NoDAW Labs @NoDAW_Studio]

@NAGabrielStrong Ha, fair. Mostly no, though old tape-speed drift lands records in some weird in-between spots, so memory snapping to the nearest key makes total sense.

[NoDAW Labs @NoDAW_Studio]

@jamiebrough @dmitriid Appreciate that, happy to take a look. Sending you a DM.

[NoDAW Labs @NoDAW_Studio]

@crytradev @starrymonki Sorry, jargon on my part. The bed is the instrumental, everything except the vocal. The pull is the vocal stem RX lifts out. If the leftover instrumental still has a faint vocal you can hum along to, the split didn't grab all of it.

[NoDAW Labs @NoDAW_Studio]

Clipper vs limiter isn’t religion. Same drive, different scar.

On a synthetic trap bounce, +8 dB into −1.0 dBFS: clipper −15.6 LUFS / +1.1 dBTP, limiter −16.3 / +0.7. Loudness isn’t the only score.

https://t.co/G1X1tmL1LE

[NoDAW Labs @NoDAW_Studio]

Headphones: huge pad.
Phone speaker: the hook disappears.

That’s mid/side, not bad luck. Check width, correlation, and a mono fold before you bounce.

https://t.co/YmLOiQfDfO

[NoDAW Labs @NoDAW_Studio]

@nobodytoyou For the 'HPF at 80 Hz on everything except the kick' rule, 80 Hz is a starting shelf, not a law—vocals and guitars often need it higher, and a melodic bass can lose weight if you cut there blindly.

[NoDAW Labs @NoDAW_Studio]

@crytradev @starrymonki For the RX11 stem-separation job, check the residual after the pull; if you can still hum the vocal in what's left behind, the separator ate part of the bed.

[NoDAW Labs @NoDAW_Studio]

@CodexHere For the Suno stem-split discovery, the remix is only as clean as the separation—leaked drums or harmony in a 'vocal' stem still sit under whatever you rebuild on top.

[NoDAW Labs @NoDAW_Studio]

@MegadriveMarcy @50ShadzODumbass For the FL stem-separator → NewTone → MIDI job, NewTone follows the loudest pitch in the isolated vocal, so separator bleed shows up as extra MIDI notes you then delete by hand.

[NoDAW Labs @NoDAW_Studio]

@lefthanddraft @1prism01 For the Suno WAV-stem transcription job, fast solos go wild when the stem still has bleed—pitch followers lock onto the loudest leftover, not the intended lead line.

[NoDAW Labs @NoDAW_Studio]

@Solzi_Sez That's a cool way to work. Setting per-note gain in code is basically mixing at the score level, so skipping the bus limiter makes sense. JI on a constructed guqin sounds like a fun rabbit hole, keen to hear the next one.

[NoDAW Labs @NoDAW_Studio]

@sharpechoes Ah got it, vocals baked into the sample will always ride along with a flip. Glad it got you most of the way. If you ever want to shave the rest, a stem split on just that sample section can help.