Problem in cloning the repo
when i was trying to clone the repo , i tried both token and ssh to clone the repo but it asks for username and password when we run the clone command and the username and pass has been changed from huggingface so it doesnt work, any other way i can clone the repo??
this was musicgen small I moved onto current AI 3.0 cuda 12.6/12.8
im rerealasing it soon
https://www.reddit.com/r/StarWarsTheoriesAI/
you can find the demo MCP reddit archive here for music ones
this was built on INT 8 originally for low end AI it was to hard I moved onto large 5000 12.6 cuda 16-24 gb required
I found INT 8 really hard to work with music it does work but its so clunky re mathing points and transformers
Sorry frequency timing + point + float + parameters+ music is very difficult this was my 1st SDK build I moved onto a more complex proper build
I also bought better rigs to run this as well since then
Ill update this thread when im done if you got a low end id rebuild this if you can its mostly there
Theres quite a bit of transfoirmer point etc float changing it uses all three I think I forget music small medium
i was able to use your code by manually downloading the files and modifying the sh file a lil bit, just one more question in other one of your repo's it says full song generation, is that doable?
thanks for the info though.
I am working on that hold on its SING SONG your referring to at the time of this build I barely read the docs I learned since then updates CUDA 12.1 SM 80 3000 series 12 gb works good do not go over 4 series might work 5 series needs a driver compile (takes 6-18 hours not fun ><) 12.1 11.8 rdy out of box
its as difficult as full music gen float 16/32
works in tandom I also have the clint app and api up
MCP rdy
will update soon for this build SING SONG is the SDK your reffering too this as far a sive gotten
Current lab MCP switching to more complex AI sing song etc vectoring etc planar CV etc etc
okay bro thanks for all the info, GG
okay bro thanks for all the info, GG
NP check back in a few weeks im in test build out ill update when I add audio background wastland musicgen and samples TTYL
ill release a build and prob drivers so 5 series doesnt have to compile i just gotta figure out how export my GCC build
brooooo, the vid looks sick to be generated by a prompt, if im not wrong
brooooo, the vid looks sick to be generated by a prompt, if im not wrong
im working on it still that wa sbeta alot firther along now
got folly background sounds explosions etc etc etc that was just my original build
still in private beta not released yet
As well I only added sound I have to ad d my open source VEO 3 ai trained from comfy to LLM as well it will be full motion audio
changing AI audio gen to fallout
still got some work to do about half way done
Curre
nt working model gotta recode pony images working on emotion and sound timing etc etc etc variablesUPDATED sm180 full build out and API build no MCP RAG yet
🚀 UPDATED: SM80 Full Build & API Out! No MCP RAG yet—optimized for 3000-series GPUs. Grab it: https://huggingface.co/ghostai1/GHOSTSONAFB 🎵






























