Movie Diffusion is SD 1.5 trained on 35mm film stills from popular movies.
Get a custom captioned model or donate towards the compute for the training of even larger models in the future:
https://ko-fi.com/dalle2pictures
Token : 35mmstyle
The visual look of 35mm film is often considered superior to other formats due to its ability to capture rich, nuanced details with a natural, organic feel. Compared to digital cameras or smaller film formats, 35mm film produces a unique texture and grain structure that is difficult to replicate with modern technology, so I fine-tuned a model to replicate the 35mm style.
Base Sd doesn't fully understand the concept of 35mm, so I finetuned a model + captions.
Add the word "photo" for most photorealism. Yes, it can more than just the face (*UPDATE : Version 1.2 was trained on more captioned full body shots / NSFW film stills (with more steps)
*Note : may have to use some light negative prompting to get your desired result, but most of the time, it will give you what you're looking for.
*2/11/2023 : v1.0 released
*2/13/2023 : v1.2 released
*2/13/2023 : Safetensors added
This model was trained on 512x512px images but can work with higher resolutions.
Let me know what improvements can be made, Enjoy!
@Dalle2Pictures (Twitter)
Description
Add the word "photo" for most photorealism. Yes, it can more than just the face (*UPDATE : safetensors version)
Token : 35mmstyle
*Note : may have to use some light negative prompting to get your desired result, but most of the time, it will give you what you're looking for.
FAQ
Comments (5)
Diffusers version please?
hey, dude, could you do a Lora on this?
nice
Could you add this checkpoint on Tensor Art, please? (https://tensor.art) Thank you in advance!
Such a good model. I still prefer to use this in supermegabadass video upscaler to better blend my latent masking bodyswaps into the original background.