The Stoked Reality models have been my most successful to date and I was pleased with v2.0: the textures, prompting response, and it was another step closer to realism than previous versions. Then when coupled with the Stoked Reality LORA I felt the colours were richer and skin details were more varied and defined.
Read more: StableDiffusion: Stoked Reality v2.1 Released
Changes in v2.1
- Integrated the Stoked Reality v1.0 LORA, along with some others I haven’t released yet,
- Updated models used in the merge
- Colours are brighter and more vibrant
- Realism of textures and details
- Improved prompting response
- Usually good hands without too much bother
- Good results with no duplication upto just under 1000pxl * 1000pxl
Whilst this model is still based on StableDiffusion 1.5 and not SDXL excellent results can still be achieved approaching those of SDXL with minimal duplication at around 960² pixels.
- Stoked Reality v2.1 fp32 (safetensors)
- Stoked Reality v2.1 fp16 (safetensors)
- Stoked Reality v2.1 fp32 InPaint (safetensors)
- Stoked Reality v2.1 fp16 InPaint (safetensors)
- vae-ft-mse-840000-ema-pruned.pt (VAE is baked into the model but I included it anyway)
This model is currently only available on CivitAI.com so why not have a look at some other models and resources while you are there.