Type | |
Stats | 2,008 |
Reviews | (56) |
Published | Mar 14, 2023 |
Base Model | |
Trigger Words | rz884n4l0g |
Hash | AutoV2 C9F0CAFEDF |
2.1 Lora Model for Analog photography
Please see in the versionning BELOW for 1.5 download link.
What about a chemical processed diffusion ?
Analog Ext is available on Gumroad as pay what you want, featuring exclusives assets with different finetuned version and his noise offset variant, an experimental 840*840 training, and a light leak helper embedding as long a 18 nice prompts !
Mostly trained on shots using Cinestill 800t films, pushing your analog generations to the closest reality level.
Please use the forAnalogPortrait negative embeddings (found in versioning below the main embedding), it will cleanse unwanted styles to try to perform only analog outputs, please note that this negative is very strict.
Note : demo pictures was generated from raw 2.1 non-ema without VAE loaded.
If you want to generate landscapes who looks very natural, be kind on negatives, analog grain / glitches can be considered as “bad” for popular negative token.
512*768 / 640*936 for portrait are safe
For achieving perfection, you can also find your base subject / scene with no negatives at all then adding some gradually.
An rz884n4l0g photo of a <subject>
An rz884n4l0g 80’s photography of …
An rz884n4l0g movie scene of
For a lighter effect, removing the trigger works fine and still keeps the major bias, increase weight if needed.
1.5 notes
Use under 0.6 strength if you are generating with a 512*512 resolution
Trained on a 768*768 dataset, quite hard to get some grain, authentic analog effect on comparison of 2.1.
Troubleshooting
If you are getting the error "failed to match keys when loading Lora" it means you are either on the native Lora integration of Automatic1111 which may not support LoCon, please install the https://github.com/kohya-ss/sd-webui-additional-networks extension to solve the issue.
Happy tuning everyone !
romerorz.art - Twitter - DeviantArt - Gumroad