True Real Vision All In One
- Includes Clip LfullFP32Zer0int_textImprovedFP32 and t5xxl_fp8_e4m3fn
This is the flux version of the True Real Vision model, designed for enhanced Realism and NSFW image generation. It outperforms the base flux dev model by integrating multiple LoRAs fused into a final checkpoint.
Settings
Steps: 8-40 (Low steps may struggle with full body anatomy)
Guidance: 3.5 - Use a higher guidance if image is not following prompt or NSFW
Sampler: dpmpp_2s_a Most sampler's work see grid for examples
Scheduler: Beta, sgm_uniform, ddim_uniform(Desaturated images)
VAE: Automatic
resolution: 768x1280
Sampler dpm2 with Scheduler Karras or exponential also work. Is good for a second pass Upscale.
Try my model in ComfyUI! TrueRealVision Flux | ComfyUI Workflow |