In my workflow i have a number of "timbre-ref" audio files that i use daily, alternating them on the same song text or on the same audio cover, to have a defined and coherent set of variants.
When they start being more than 5/6 , the songlist can become easily congested so i've modded the web interface as follows :
- added a checkbox in the song list to mark the audio file (with decoded VAE) as "timbre-palette"; the song will be hidden from the songlist and added to the "timbre dropdown" menu ,right under the "Task" submenu.
- added a "timbre reset" button under the timbre dropdown menu that reset the song state, removing it from the "palette" pool and unhiding it in the songlist.
For my exigences it a "quasi mandatory" change but i understand that it's probably relevant to a really small subset of people so i've not posted any pull request, but if you are interested just say it...
Thanks for all the work. Acestep.cpp is working flawlessy on my Maxwell based GPU with 4Gb Vram, 16Gb system Ram, i76700HQ on Linux 6.12: time to render with sftTurbo50XL DiT is +-1.60 x (length in seconds of the track); +-0.90 x (length) with SftTurbo50 dit , 10 steps for both.
Pillopaus
In my workflow i have a number of "timbre-ref" audio files that i use daily, alternating them on the same song text or on the same audio cover, to have a defined and coherent set of variants.
When they start being more than 5/6 , the songlist can become easily congested so i've modded the web interface as follows :
For my exigences it a "quasi mandatory" change but i understand that it's probably relevant to a really small subset of people so i've not posted any pull request, but if you are interested just say it...
Thanks for all the work. Acestep.cpp is working flawlessy on my Maxwell based GPU with 4Gb Vram, 16Gb system Ram, i76700HQ on Linux 6.12: time to render with sftTurbo50XL DiT is +-1.60 x (length in seconds of the track); +-0.90 x (length) with SftTurbo50 dit , 10 steps for both.
Pillopaus