Models ‐ checkpoints, LoRAs and inbox - runew0lf/RuinedFooocus GitHub Wiki
Checkpoints and LoRAs
inbox
There is a special folder models\inbox\
. When RF starts, the hash of the files in this folder will be checked with CivitAI and then moved to a sub-folder in checkpoints
or loras
. For example, SDXL models will be moved into checkpoints\SDXL 1.0
and so on. A nice feature if you are lazy or if you aren't sure where to put your model.
Previews
Ruined Fooocus will try to find previews for your models on CivitAI. If you want to change a preview or add your own without getting it from CivitAI, just place an image named the same as the model in cache\checkpoints
or cache\loras
. RF will also save information about the model in a .json
file and LoRAs will have their trigger words written to a .txt
file.
Animated preview on CivitAI will be saved as a .gif.
There are also some settings you can set in settings\settings.json
:
"model_preview": "zoom,caption,nogifzoom",
- zoom - will scale the image to better fit the model select boxes on the webpage, and also save space.
- caption - will print the name of the model in the upper left corner of the image. Best used together with
zoom
. - nogifzoom - By default, animations will be scaled down to save space. If you really want huge gifs you can add
nogifzoom
(not recommended)
These will only affect images when they are downloaded, to update an image from CivitAI, simply remove it and let RuinedFooocus download it again.
Example:
path.json
path_checkpoints
and path_loras
can either be a string pointing to the folder where your models are stored or a list of strings, allowing you to have multiple folders with models. (If you have multiple folders, models in inbox
will be moved to the one first in the list.)
path.json
"path_checkpoints": [
"..\models\checkpoints",
"D:\ComfyUI\models\checkpoints"
],
"path_loras": [
"..\models\loras",
"D:\ComfyUI\models\loras"
],
"path_inbox": "C:\Downloads\inbox",