Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
This is my "Ultimate" SDXL workflow. After about a year of messing around with SDXL checkpoints. This is what i've sort of landed with that has a good balance of quality and speed. 3 minutes from 0 to 2721x6167 that...
Runtime profile
Source description
Intro
This is my "Ultimate" SDXL workflow. After about a year of messing around with SDXL checkpoints. This is what i've sort of landed with that has a good balance of quality and speed. 3 minutes from 0 to 2721x6167 that blows the doors off Flux. It tries to push SDXL to the max in terms of detail and resolution trying to match FLUX. Tons of great SDXL models that I wish had the clean Flux look. You can make stuff that looks VERY close to Flux by using alot of these great extentions that in my opinion put SDXL on Meth. Some of this stuff is very obscure and most workflows are overly bloated with nonsense. Hopefully this is something that you can just expand on as a base.
QuickStart
Everything is initially setup for pure text-to-image. Enter a prompt and hit queue. Example prompt is provided.
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
SDXL Ultimate 16 MegaPixel is an SDXL workflow for generating images from text prompts, refining source images through image-to-image or depth-controlnet guidance, and upscaling results through tiled stages toward 8MP and 16MP outputs.
The default path is pure text-to-image: enter a prompt and queue it. Enable the image-to-image or depth-controlnet group for those paths.
The notes describe composition and cleanup around 2MP before tiled refinement at 8MP and 16MP.
Inputs are text prompts, source images for image-to-image, and optional depth-controlnet guidance.
The workflow covers text-to-image, image-to-image, and upscaling, with stages described at 2MP, 8MP, and 16MP.
It uses SDXL 1.0 as its base model and is published by Smeks.
Named model files are 4x_foolhardy_Remacri.pth, longclip-L.pt, sdxl_lustify_endgame.safetensors, and sdxl_tile_realistic_v2_fp16.safetensors.
Named workflow components include ACN_AdvancedControlNetApply_v2, ACN_ScaledSoftControlNetWeights, ACN_TimestepKeyframeInterpolation, Any Switch (rgthree), Anything Everywhere, Anything Everywhere3, ApplyFBCacheOnModel, BasicGuider, BasicScheduler, CheckpointLoaderSimple, CLIPTÉ
Other named workflow components include CLIPTextEncodeBREAK, ControlNetLoader, DetailDaemonSamplerNode, DynSamplerSelect, easy imageBatchToImageList, easy imageListToImageBatch, Empty Latent by Pixels (WLSH), Fast Groups Muter (rgthree), GetNode, Image Comparer (rgthree), Image
SeaArtLongXLClipMerge is described as extending CLIP handling from 77 to 248 tokens.
Conditional merges are entered in one prompt with BREAK separators; the notes describe PerturbedAttention settings of 1.5–2.
For the 8MP and 16MP stages, the notes specify 30% denoise for defogging and sharpening.
Before use, verify that 4x_foolhardy_Remacri.pth, longclip-L.pt, sdxl_lustify_endgame.safetensors, and sdxl_tile_realistic_v2_fp16.safetensors are available; no source URL is provided for them.
Suggestion · not verified
Verify that the listed workflow components are available in your setup before use; no source URLs are provided for them.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
1 excerptSource context: 703 downloads · Type Workflows · Base model SDXL 1.0
Optional Stuff
Everything is setup to do depth-controlnet or image-to-image without messing with anything. Just enable the following group and go.
image-to-image
depth-controlnet
Advanced
There are many knobs to tweak. Honestly, you'll have to experiment and see what does what, but breifely I think its worth at least noting from my experience most important bits.
SeaArtLongXLClipMerge All SD checkpoints have a clip limited to 77 tokens and so this patches any clip to 248 tokens allowing alot more detailes. Actual human readable prompts work with it much better similar to flux but it's still very limited. Each tile can hold 248 tokens generated from Florence + base prompt which gives unbelevable results at high tile counts.
Clip Text Encode with BREAK This in conjuction with the long clip merge node gives alot of new possibilities and simplifies a ton of noodles around conditioning. Conditional merges are all in one prompt seperated by the BREAK.
CLIP NegPip + PerturbedAttentionGuidance This definetly helps cleanup the image alot. Generations are much more coherent and make sense. Alot of artifacting goes away by simply having these enabled. A setting of 1.5-2 is about all you really want.
Dynamic Sampler + Detail Daemon I have all this setup how i want but you may want to experiment, but these dynamic samplers push alot more refined + smooth details out of the image almost like flux. Honestly you dont even need to pass the image through Flux because it looks that clean.
Conclusion
Basically steps 1 and 2 are the primers for the actual tiling steps 3 and 4. The image has to be nearly perfect in composition and mostly free from any major artifacting before going into tiling. Step 1 tries to just get a good composition and you may have to change around conditioning and CFG ramping to get a desired result depending on model used etc. Step 2 does the heavy lifting to clean up everything and 2 Mega Pixels seems to be the sweet spot with the TTPlanet Realistic Tile Controlnet before requiring full breakup into tiles. Step 3 and 4 require a nearly perfect image for them to work properly. Going to 8MP and 16MP only requires 30% denoise for "defogging" and "sharpening" everything to bring it all into focus. After 2MP the need for CFG and denoise goes away as you already have the image. It's really just a matter of massaging defects out along the way to 16MP. Thats just my perspective on it all.
Curious to see what people think and would wanna see anything that could improve on this because ive tried so many workflows in the past and nothing ever seemed to really go as far as this.
Estimated VRAM requirement
Estimate unavailable
63.9 MB across 1 of 8 model files. Model file total + 25% loading overhead + 2 GB execution buffer, rounded up.
Requirements
50 requirementsUpscaler · 63.9 MB · PTH · Unknown
ImageUpscaleWithModel
Not resolvedUpscaler · Unknown
longclip-L.pt
Not resolvedText encoder · Unknown
sdxl_lustify_endgame.safetensors
Not resolvedCheckpoint · Unknown
sdxl_tile_realistic_v2_fp16.safetensors
Not resolvedControlNet · Unknown
UpscaleModelLoader
Not resolvedUpscaler · Unknown
VAEDecode
Not resolvedVAE · Unknown
VAEEncode
Not resolvedVAE · Unknown
ACN_AdvancedControlNetApply_v2
Not resolvedNode pack · Unknown
ACN_ScaledSoftControlNetWeights
Not resolvedNode pack · Unknown
ACN_TimestepKeyframeInterpolation
Not resolvedNode pack · Unknown
Node pack · Unknown
Anything Everywhere
Not resolvedNode pack · Unknown
Anything Everywhere3
Not resolvedNode pack · Unknown
ApplyFBCacheOnModel
Not resolvedNode pack · Unknown
BasicGuider
Not resolvedNode pack · Unknown
BasicScheduler
Not resolvedNode pack · Unknown
CheckpointLoaderSimple
Not resolvedNode pack · Unknown
CLIPTextEncodeBREAK
Not resolvedNode pack · Unknown
ControlNetLoader
Not resolvedNode pack · Unknown
DetailDaemonSamplerNode
Not resolvedNode pack · Unknown
DynSamplerSelect
Not resolvedNode pack · Unknown
easy imageBatchToImageList
Not resolvedNode pack · Unknown
easy imageListToImageBatch
Not resolvedNode pack · Unknown
Empty Latent by Pixels (WLSH)
Not resolvedNode pack · Unknown
Fast Groups Muter (rgthree)
Not resolvedNode pack · Unknown
GetNode
Not resolvedNode pack · Unknown
Image Comparer (rgthree)
Not resolvedNode pack · Unknown
Image Save
Not resolvedNode pack · Unknown
ImageScaleToTotalPixels
Not resolvedNode pack · Unknown
JoinStrings
Not resolvedNode pack · Unknown
KSamplerSelect
Not resolvedNode pack · Unknown
LayerMask: LoadFlorence2Model
Not resolvedNode pack · Unknown
LayerUtility: Florence2Image2Prompt
Not resolvedNode pack · Unknown
MarkdownNote
Not resolvedNode pack · Unknown
PerturbedAttention
Not resolvedNode pack · Unknown
Power Lora Loader (rgthree)
Not resolvedNode pack · Unknown
PreviewImage
Not resolvedNode pack · Unknown
RandomNoise
Not resolvedNode pack · Unknown
SamplerCustomAdvanced
Not resolvedNode pack · Unknown
ScheduledPerpNegCFGGuider //Inspire
Not resolvedNode pack · Unknown
SeaArtLongXLClipMerge
Not resolvedNode pack · Unknown
Seed Everywhere
Not resolvedNode pack · Unknown
SetNode
Not resolvedNode pack · Unknown
ShowText|pysssss
Not resolvedNode pack · Unknown
Text Multiline
Not resolvedNode pack · Unknown
TiledDiffusion
Not resolvedNode pack · Unknown
TTP_Image_Assy
Not resolvedNode pack · Unknown
TTP_Image_Tile_Batch
Not resolvedNode pack · Unknown
TTP_Tile_image_size
Not resolvedNode pack · Unknown
