No custom nodes needed! This is a very simple workflow for merging SDXL models by blocks/layers. You can load it, select 2 models, adjust the merge ratio, and you can start generating
Perfil de ejecución
Descripción de la fuente
No custom nodes needed! This is a very simple workflow for merging SDXL models by blocks/layers. You can load it, select 2 models, adjust the merge ratio, and you can start generating
As general info: IN blocks: contain concepts/composition/etc. OUT blocks: contain style/aesthetics The further away the blocks are from the middle, the closer they are to pixel level
SDXL has 20 blocks while SD1.5 has 26 blocks! Don't confuse them since most guides might show 26 blocks or IN11/OUT11 like in the image below. The merging logic stays the same though.
Here is an image from this article that helped me understand different blocks: Here is another great article by RestlessDiffusion giving a tutorial on block merging in A1111/Forge using the SuperMerger extension and showing the effects of different merge ratios:
Comentario generado por IA
Explicación generada por IA basada en los detalles de la fuente y la configuración. Las sugerencias están claramente identificadas.
Use this ComfyUI workflow to merge two SDXL models by blocks or layers, adjust the merge ratio, and generate images.
Load the workflow, select two models, adjust the merge ratio, and start generating.
The main inputs are two SDXL models and a merge ratio.
The workflow includes image-preview and image-saving steps named PreviewImage and SaveImage.
The workflow targets SDXL 1.0, and its description says no custom nodes are needed.
Listed node names include CheckpointLoaderSimple, CheckpointSave, CLIPMergeSimple, CLIPTextEncode, EmptyLatentImage, and KSamplerAdvanced.
Other listed node names are ModelMergeSDXL, PreviewImage, SaveImage, VAEDecode, and VAELoader.
Listed model entries are 4thTailAnimeHentai_050.safetensors, 4thTailAnimeHentai_v045.safetensors, ponyStandardVAE_v10.safetensors, and checkpoints/ComfyUI.
IN blocks are described as containing concepts and composition, while OUT blocks are described as containing style and aesthetics.
Blocks farther from the middle are described as closer to the pixel level; SDXL has 20 blocks, while SD1.5 has 26.
When following a block-merging guide, check whether its numbering refers to SDXL's 20 blocks or SD1.5's 26.
Sugerencia · sin verificar
Before using the workflow, verify that the listed model entries are available in your ComfyUI setup.
Sugerencia · sin verificar
¿Es necesario editarlo?
Inicia sesión para solicitar una edición.
Fuentes
1 fuenteExtractos de fuentes
1 extractoSource context: 489 downloads · Type Workflows · Base model SDXL 1.0
Requisito de VRAM estimado
Estimación no disponible
Los tamaños de los modelos aún no están lo bastante completos para calcular este requisito.
Requisitos
15 requisitos4thTailAnimeHentai_050.safetensors
No resueltoCheckpoint · Unknown
4thTailAnimeHentai_v045.safetensors
No resueltoCheckpoint · Unknown
Checkpoint · Hugging Face · checkpoints/ComfyUI
ponyStandardVAE_v10.safetensors
No resueltoVAE · Unknown
VAEDecode
PosibleVAE · Unknown
VAELoader
PosibleVAE · Unknown
CheckpointLoaderSimple
PosiblePaquete de nodos · Unknown
CheckpointSave
PosiblePaquete de nodos · Unknown
CLIPMergeSimple
PosiblePaquete de nodos · Unknown
Paquete de nodos · Unknown
EmptyLatentImage
PosiblePaquete de nodos · Unknown
KSamplerAdvanced
PosiblePaquete de nodos · Unknown
Paquete de nodos · Unknown
PreviewImage
PosiblePaquete de nodos · Unknown
SaveImage
PosiblePaquete de nodos · Unknown