Module modifier

Add additional resources to preset::Preset

Functions

• hybrid_taesd
  Apply https://huggingface.co/cqyan/hybrid-sd-tinyvae taesd autoencoder for faster decoding (SD v1/v2)
• hybrid_taesd_xl
  Apply https://huggingface.co/cqyan/hybrid-sd-tinyvae-xl taesd autoencoder for faster decoding (SDXL)
• lcm_lora_sd_1_5
  Apply https://huggingface.co/latent-consistency/lcm-lora-sdv1-5 to reduce inference steps for SD v1 between 2-8 cfg_scale 1. 4 steps.
• lcm_lora_sdxl_base_1_0
  Apply https://huggingface.co/latent-consistency/lcm-lora-sdxl to reduce inference steps for SD v1 between 2-8 (default 8) Enabled [api::SampleMethod::LCM]. cfg_scale 2. 8 steps.
• real_esrgan_x4plus_anime_6_b
  Add the https://huggingface.co/ximso/RealESRGAN_x4plus_anime_6B upscaler
• sdxl_vae_fp16_fix
  Apply https://huggingface.co/madebyollin/sdxl-vae-fp16-fix to avoid black images with xl models
• t5xxl_fp8_flux_1
  Apply https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp8_e4m3fn.safetensors Fp8 t5xxl text encoder to reduce memory usage
• t5xxl_fp16_flux_1
  Apply https://huggingface.co/comfyanonymous/flux_text_encoders/blob/main/t5xxl_fp16.safetensors Default for flux_1_dev/schnell
• taesd
  Apply https://huggingface.co/madebyollin/taesd taesd autoencoder for faster decoding (SD v1/v2)
• taesd_xl
  Apply https://huggingface.co/madebyollin/taesdxl taesd autoencoder for faster decoding (SDXL)