Search Results
6/16/2025, 1:57:38 AM
>>101219905
I use VACE for all the ones in that post, that one used this video as reference. Konata was a first frame last frame, and subaru was loli-kami.
https://gelbooru.com/index.php?page=post&s=view&id=12001782&tags=webm+pyra_%28xenoblade%29+
Oh yeah there are a lot out there now, I mostly use the VACE model (the 14B one), it to me is the most intuitive since you have actual control over the video rather than pray to rng gods kek. In general though I think the ones you want to use are 14B I2V models since you don't have to worry about style+character LoRAs just action ones (for coom mostly, it can sfw stuff great without). Any one of them is fine (except the fusion x one, that one is mostly for t2v).
If you want to use VACE, I think this workflow shows off it's capability, Though I think you make pretty good workflows so probably can make better than this guy (his is hard to follow, not that mine is any better lol)
https://civitai.com/models/1680850/wan21-vace-14b-t2v-i2v-v2v-controlnet-masking-simple-comfyui-workflow?modelVersionId=1902439
GGUF for vace are here
https://huggingface.co/QuantStack/Wan2.1_14B_VACE-GGUF/tree/main
>2nd GPU
I don't know about a 2nd gpu that's really if you feel like it's worth your money unless you mean in terms of split computing than I think it doesn't allow multi-gpu rendering though I am not a 100%. You definitely need more than 12gb of vram though, I am just barely scraping by, one misclick and boom my computer hangs....
I use VACE for all the ones in that post, that one used this video as reference. Konata was a first frame last frame, and subaru was loli-kami.
https://gelbooru.com/index.php?page=post&s=view&id=12001782&tags=webm+pyra_%28xenoblade%29+
Oh yeah there are a lot out there now, I mostly use the VACE model (the 14B one), it to me is the most intuitive since you have actual control over the video rather than pray to rng gods kek. In general though I think the ones you want to use are 14B I2V models since you don't have to worry about style+character LoRAs just action ones (for coom mostly, it can sfw stuff great without). Any one of them is fine (except the fusion x one, that one is mostly for t2v).
If you want to use VACE, I think this workflow shows off it's capability, Though I think you make pretty good workflows so probably can make better than this guy (his is hard to follow, not that mine is any better lol)
https://civitai.com/models/1680850/wan21-vace-14b-t2v-i2v-v2v-controlnet-masking-simple-comfyui-workflow?modelVersionId=1902439
GGUF for vace are here
https://huggingface.co/QuantStack/Wan2.1_14B_VACE-GGUF/tree/main
>2nd GPU
I don't know about a 2nd gpu that's really if you feel like it's worth your money unless you mean in terms of split computing than I think it doesn't allow multi-gpu rendering though I am not a 100%. You definitely need more than 12gb of vram though, I am just barely scraping by, one misclick and boom my computer hangs....
Page 1