Downloads · 30 days
28
1% of all-time downloads
romcom-diffusion-1.0
jkcarney/romcom-diffusion-1.0
romcom-diffusion-1.0 is a text-to-image model from jkcarney. Use it when you need an image from a text prompt. It is set up for diffusers. The card lists the license as creativeml-openrail-m.
All-time downloads
3.7K
Public
Repo size
14 GB
Likes
12
Public
Repo makeup
Click a slice to open those files.
.bin5.5 GB · 39%
At a glance
- Task
- Text-to-Image
- Library
- diffusers
- License
- creativeml-openrail-m
- Access
- Public
- Created
- Dec 19, 2022
- Updated
- Jan 30, 2023
- SHA
- eaba76e8
Model card
From the Hugging Face model README
romcom-diffusion is a Dreambooth model fine tuned on cropped 512x512 stills from various 1990s era romantic comedy movies. These stills are primarily of human subjects, but there are also samples of wide shots, crowds, and cars. Currently the list of movies used is:
- 10 Things I Hate About You (64)
- Clueless (45)
- American Pie (37)
In the future I may release a new version with more movie stills from additional movies. My hope is that more samples from more movies would increase facial fidelity and style.
In the prompt, use activation token: romcom style
Trained using Stable Diffusion 1.5 as a base.
I also recommend including movie still in your prompt; it was the class prompt for regularization images and I found it produces the best results.
I've found the best results using the DPM++ 2S a Karras sampler as well, but obviously feel free to experiment.
Sample image generation parameters can be found in prompts.txt
- Task
- Text-to-Image
- Library
- diffusers
- License
- creativeml-openrail-m
- Languages
- en
- Created
- Dec 19, 2022
- Updated
- Jan 30, 2023