This LoRA is designed to capture the essence of 20s Silent movies.
The trigger words are ArsMovieStill, 20s Silent Movie Still
Good keywords Black and White, monochrome, greyscale, sepia
Description
FAQ
Comments (14)
it looks amazing! specially the SDXL one is on point to "the real thing". Thanks.
Thanks :) Pony is just plain bad at "Realistic" images (or any complex concept) - I honestly considered droping it from my Training, but too many people seem to like it, even with the limitations it has. And SD 1.5 is just OLD at this point, but I don't want the people with low end video to feel left out.
@ArsMachina oh c'mon, sd 1.5 it's really not that old, isnt? it's low res that much is true but it still is very useful for limited hardware. SD 1.5 is really mindblowing for its size. About Pony... i don't use those, so i don't know. But its nice they can have their own version, so very well done. btw, Your 70's horror lora is looking really good. I love the 70's aesthetics.
@ElectricDreams True :) To be honest I still use SD 1.5 for my more complex Workflows as my current Video is only 8GB. SDXL with 3-4 Control Nets + Relight + innitswap+ other goodies need 12/16GB+ .
Basically most of the times, if you see sd 1.5 version of a LoRA , this means that I needed it for something more complex :)
@ArsMachina yeah, i'm on the same boat. You can make a lora in 14 minutes with OT too (compared with the 2-3 hours of SDXL). So, yes, pretty useful. Looking forward to Nvidia's Sana Model to have a good alternative to SD 1.5.
Great lora love it.
SDXL lora runs like a dream, I seriously love this one.
Unfortunately I can't get the FLUX version to run.
I keep getting a string of "lora key not loaded: diffusion_model.double_blocks.1.txt_attn.qkv.lora_down.weight
lora key not loaded: diffusion_model.double_blocks.1.txt_attn.qkv.lora_up.weight
... etc"
errors.
After looking closer I noticed that Comfy 'sees' it as an SDXL Lora thus incompatible with FLUX.
Any suggestions would be gratefully appreciated!
Everytime I get a strange error in ComfyUI that blocks my work I mass update everything. Nothing special about this one , so not sure what is causing your issue. The only thing I can think off is that I trained it here on Civit and have not changed the names of the different versions, so when you download them they will have the same one - make sure you are not actually selecting SDXL version by mistake.
And thank you for the kind words :) I REALLY enjoy all of the Movie Stills LoRAs and currently am training 2 new ones (90s inde and update to the 80s) and in the process of creating the dataset of a 3rd one :)
@ArsMachina Thanks for the reply!
Yeah I have all my models/Loras in there respective folders. SDXL -> Loras/Sdxl/, FLUX -> Loras/Flux/ and so on. I re-downloaded the Flux model AND renamed it to be safe but still get the same error.
Everything up to date.
Anyway, I'm having a blast with the xdsl and pony variants. So much appreciated!
P.s. can't wait to get my hands on you next batch!
Just a quick update,
Well, I re downloaded it again today just for the hell of it and IT WORKS!!! So it must have been a corrupt download on my end!
Sorry for the hassle!
So good, this one is a lot of fun.
Excellent LoRA, right on target! Great job all over. Wish I'd seen this one before, would have saved me some effort! XD
Really nice collection you have curated there as well :)
Thank you, you're very kind ! Glad you liked it.