thecollabagepatch
@thepatch_kev
musician accidentally becomes chad @_buildspace s5 grad
ever since i saw @jesseengel 's audio injection demo for magenta-rt, i've been down a rabbit hole with this looper app. create loops with jerry (stable-audio-open-small). send them to magenta as combined audio to 'jam' on. very 'vibe coding for music' now....🧵
when docker hub goes down right as you're finishing up building torch from source
jerry (aka stable-audio-open-small) is part of all my UIs now finally got a homie in my discord server aggressively finetuning the base model in various ways so i added a model switcher for any SAOS checkpoint on hf keeps the arc trained jerry loaded in cache and switches out…
it sounds counterintuitive, but the more i use neural networks in my music production the more i focus on getting better at playing irl instruments. better input audio makes for more dank model outputs and i can learn what the model did by ear and play it.
so there's this thing sora has been able to do since v1 apparently where you can give it input video and have it extend the video backwards in time wondering why we don't have this for music models... does it exist and i just dunno about it? @zqevans @zacknovack
bedroom studios are awesome until the neighbor's dog won't stop barking
Introducing AppRejection Apple works hard to reject your app and waste your days. AppRejection uses AI agents to mock review your app in minutes. We also include a free git compliance tool. Reply with "REJECTED" for private beta link + 50% code
codex is finally cleaning up some functional but unreadable solo-coder slop for me i just hope i can still read it when its done
building something for local usage without having the gpu to run it locally brings a specific kind of stress to development
making an api is only fun if you use it in several places using @GoogleMagenta 's magenta-rt, currently just doing single-shot generations of 4/8 bar loops using the DAW (or other model outputs) as audio context SAOS: lofi hiphop with pianos -> magenta: saxophone updates otw
ever since i saw @jesseengel 's audio injection demo for magenta-rt, i've been down a rabbit hole with this looper app. create loops with jerry (stable-audio-open-small). send them to magenta as combined audio to 'jam' on. very 'vibe coding for music' now....🧵
i might be the only person who uses LLMs to code without writing tests... they just feel like an excellent way for me and the robot to gaslight ourselves and then spend a day debugging the test.
UncomfortableUI
darius in the daw day one trying to reach ozone levels of resource intesiveness with this thing
darius in the daw day one trying to reach ozone levels of resource intesiveness with this thing
i feel like we're finally starting to get to a place where music models become more like instruments we play. we just need to make them smaller
as i writer by training i truly don't mind typing this much when im coding
might get to learn how stone age my llm workflow is at 3pm should be good times
hi i worked on this all day now come to my party tomorrow if ur shipping, lrn2ship faster if ur investing, find ur next moonshot if ur something else, come make us care c u in the ring the legend never dies
United States Trends
- 1. South Carolina 31K posts
- 2. #EubankBenn2 21.7K posts
- 3. Texas A&M 30.3K posts
- 4. Bama 9,888 posts
- 5. Beamer 8,863 posts
- 6. Ryan Williams 1,436 posts
- 7. Ty Simpson 1,999 posts
- 8. Heisman 8,534 posts
- 9. Makai Lemon N/A
- 10. Aggies 8,885 posts
- 11. Michigan 44.3K posts
- 12. Daniel Hill N/A
- 13. Mateer 1,755 posts
- 14. Oklahoma 19.1K posts
- 15. Sellers 14K posts
- 16. Northwestern 7,524 posts
- 17. Gio Reyna N/A
- 18. Tate Sandell N/A
- 19. Cuevas 6,348 posts
- 20. Marcel Reed 5,370 posts
Something went wrong.
Something went wrong.