Dick Gumshoe is coming XL and in a 1.5 diffusion near you

This LoRA can generate many style for our good ol' pal Dick Gumshoe. It's even able to create photorealistic cosplay. Both SFW and NSFW friendly, however the SDXL version will probably need other LoRA until more NSFW friendly checkpoints are released.

Using the term "dick_gumshoe" in the positive prompt will help tremendously in the generation of your creation. I avoided using watermarks/signatures in the dataset, however some had to be included. Should they appear, use "text, watermark" in the negative prompt.

Square sized pictures usually generate a half-body, sometime 3/4, headshot, while a portrait sized picture usually generate a full-body shot. Landscape picture tend to create multiple Gumshoe, for the better or worst.

My testing of the XL version is quite limited due to SDXL only working on CPU mode for me, but from what I have generated, it is quite good. Upscaling generally help with faces and hands, but the model is surprisingly resilient to face deformation. if generating 2D style pictures, the eyes and open mouth can cause some trouble, but nothing a face detailer/swapper can't fix.

The only real struggle is to generate the pen on his ear if not prompted. Either prompt the pen either in positive or negative prompt, or inpaint the affected area. Sometime a gun can also appears in one hand when unprompted. Use the same trick as with the pen to avoid, or improve, it.

The SD 1.5 version is now out and available. Use it with your favorite Photorealistic, Anime, 2D or NSFW checkpoint.

Recommended settings (Your mileage will vary. Checkpoints usually supersede these.):

Model Strength: 0.7-1

NSFW generations

No other LoRa is necessary to create NSFW content if the checkpoint supports it, however they do help for some position and to eliminate some questionable genital organs.