Saltar al contenido
WorkflowPublicadoExterno
Descargar

Foton's Workflow with ControlNet upscale — v1.2 — v2.1

Somewhat simple Text to Image workflow with ControlNet tiled upscale

VersiónRelease 3446275 — Foton's Gen 2.1.jsonVRAM estimadaNot yet characterizedUnratedcalidad

Perfil de ejecución

Acceso
Flujo externo
Estado
Publicado
Versión
Foton's Gen 2.1.json
Fuente
Civitai
Actualizado
3 nov 2026

Descripción de la fuente

No release description captured.

What you need

System memory
12 GiB

Remaining hardware or download-size estimates are not available yet.

Requisitos

Modelos

11
  • blue_shift_v3.safetensors

    No resueltoOpcional

    Checkpoint · Unknown

  • LoRA · 218 MB · SAFETENSORS · Unknown

  • diffusion_pytorch_model_promax.safetensors

    Archivo especificadoOpcional

    ControlNet · 2.34 GB · Hugging Face · UmeAiRT/ComfyUI-Auto-Installer-Assets

  • face_yolov8m.pt

    No resueltoOpcional

    Object Detector · Unknown

  • glinty_v1.safetensors

    No resuelto

    Checkpoint · Unknown

  • GOBOLightingILL.safetensors

    No resueltoOpcional

    LoRA · Unknown

  • hand_yolov8s.pt

    No resueltoOpcional

    Object Detector · Unknown

  • Modern_anime_render.safetensors

    No resueltoOpcional

    LoRA · Unknown

  • p4tr0nusXLP.safetensors

    No resueltoOpcional

    LoRA · Unknown

  • Upscaler · 63.8 MB · SAFETENSORS · Unknown

  • TTPLANET_Controlnet_Tile_realistic_v2_fp16.safetensors

    No resueltoOpcional

    ControlNet · Unknown

Paquetes de nodos

7
  • comfyui-impact-pack

    EncontradoOpcional

    Paquete de nodos · Registry

  • comfyui-impact-subpack

    EncontradoOpcional

    Paquete de nodos · Registry

  • comfyui-kjnodes

    EncontradoOpcional

    Paquete de nodos · Registry

  • ComfyUI-TiledDiffusion

    EncontradoOpcional

    Paquete de nodos · Registry

  • SaveImageWithMetaDataUniversal

    Encontrado

    Paquete de nodos · Registry

  • skimmed_cfg

    Encontrado

    Paquete de nodos · Registry

  • UltimateSDUpscale

    PosibleOpcional

    Paquete de nodos · Unknown

Comentario generado por IA

Explicación generada por IA basada en los detalles de la fuente y la configuración. Las sugerencias están claramente identificadas.

Propósito

  • Use this ComfyUI workflow to generate images from a text prompt and selected checkpoint, then upscale them in tiles with either Ultimate SD or ControlNet, chosen with a one-click switch.

Flujo

  • Advanced mode is the default and pairs ControlNet tile upscaling with a HiRes upscaler; Simple Mode uses Ultimate SD.

  • Optional face and hand detailers support custom prompts.

Entradas y salidas

  • Save upscaled images as PNG and/or JPG with Civitai-compatible metadata; PNGs can also contain the workflow.

Configuración inicial

  • Manually install the Save Images With Metadata Universal Extension from GitHub, and install or update custom ComfyUI nodes using ComfyUI Manager.

  • Run the Metadata Rule Scanner once after adding nodes or updating its extension, then bypass it.

Dependencias

  • Listed model files include blue_shift_v3.safetensors, DetailerILv2-000008.safetensors, diffusion_pytorch_model_promax.safetensors, face_yolov8m.pt, glinty_v1.safetensors, and GOBOLightingILL.safetensors.

  • Also listed are hand_yolov8s.pt, Modern_anime_render.safetensors, p4tr0nusXLP.safetensors, remacri_original.safetensors, and TTPLANET_Controlnet_Tile_realistic_v2_fp16.safetensors; glinty_v1.safetensors is the only model file marked non-optional.

  • The workflow lists these node packs: comfyui-impact-pack, comfyui-impact-subpack, comfyui-kjnodes, ComfyUI-TiledDiffusion, SaveImageWithMetaDataUniversal, skimmed_cfg, and UltimateSDUpscale.

Configuración

  • Promax is listed for general use, while TTplanet realistic mode is listed for realistic images.

  • For low VRAM, try a 512 tile size; if using an ESRGAN such as 4x-UltraSharp and colors look too saturated, set the color blend factor to 0.

    Sugerencia · sin verificar

Precauciones

  • The optional alternate checkpoint for upscaling is experimental and disabled by default.

  • The install notes do not name the extension that provides UltimateSDUpscale; check which extension you need before using Simple Mode.

    Sugerencia · sin verificar

¿Es necesario editarlo?

Inicia sesión para solicitar una edición.

Fuentes

1 fuente
CivitaiPrincipalVersión 3446275
civitai.com/models/2498749

Verificado 3 nov

Workflow instructions

Author notes from Foton's Gen 2.1.json · selected release

MetaData Ruler Scanner is a part of an extension for saving image metadata compatible with Civitai format. Run this subgraph once after adding new nodes to the workflow or after updating MetaData extension. You can bypass it after.

Author notes from Foton's Gen 2.1.json · selected release

Bypass CLIP Set Last Layer to disable Clip Skip

Author notes from Foton's Gen 2.1.json · selected release

CFG Skimming: allows you to crank up model CFG to higher values (>10) without blasting colors in the final output. Can enhance prompt adherense especially for long prompts (depends on the model and its CLIP). Skimming_CFG: Keep between 2 to 4 for best results when using CFG skimming. Remember to lower base CFG value to checkpoint recommended value when disabling skimming. Most common SDXL / IL / Pony Resoulutons 832 x 1216 896 x 1156 768 x 1344 1024 x 1024 CUSTOM 1024 x 1536 900 x 1600

Author notes from Foton's Gen 2.1.json · selected release

**Simple Mode** Uses Ultimate SD upscaler extension (similar to A1111 and Forge). **Recommended Simple Mode values** Denoise 0.2-0.3, lower for higher scale. Steps: 4-5 works good for 1.5-2x scale. **Advanced mode (default)** A two-step upscaling using ControlNet tile upscaling together with standart HiRes upscaler (like UltraSharp, Remacri). Requires additional model to work. Slower with some quality increase at 2x, but works better for 3x+ scale without adding typical SD upscaler artifacts especially with higher denoise value. **Recommended Simple Mode values** Denoise 0.25-0.4, lower for higher scale. My typical settings are: - for 2x scale: 0.25-0.35 denoise, 5-10 steps - for 3x scale: 0.2-0.3 denoise, 10-15 steps - for 4x scale: experiment :) **Example ControlNet models for Advanced upscaler** 1. [TTplanet for realistic models]( https://huggingface.co/TTPlanet/TTPLanet_SDXL_Controlnet_Tile_Realistic) 2. [Universal Promax model for general use]( https://huggingface.co/xinsir/controlnet-union-sdxl-1.0) Put these models into ComfyUI > models > controlnet **Additional options** - blend factor: image color correction, 0.1 is set for Remacri upscaler. Set to 0 if your upscaled image color seems off - tile_batch_size: how many tiles are processed at the same time during upscaling. By default is set to 1, use higher value if you have >12GB ram

Author notes from Foton's Gen 2.1.json · selected release

Some upscalers like Remacri reduce color contrast and saturation. Image Blend node fixes this. Bypass it if you're using a different upscaler and hires image looks too different from the original.

Download files

Artefactos del modelo

2 artefactos

fotonsWorkflowWith_v21.zip

zip · 23,0 KB · SHA-256 74e4eb6da16d…4698 · Civitai

Download file

Foton's Gen 2.1.json

json · 171 KB · SHA-256 9bd99728f041…ba1c

Download file

Extractos de fuentes

1 extracto

Source context: 581 downloads · Type Workflows · Base model Illustrious