Add FLUX3 DWPose repose workflow

This commit is contained in:
Morpheus
2026-11-07 21:07:25 +00:00
parent 950a276dca
commit c2c0cec4e5
7 changed files with 95 additions and 0 deletions
+8
View File
@@ -59,3 +59,11 @@ Use `../character_reference.png`, the canonical I2I prompt, and the installed
runtime, output dimensions, output path, visual result, and exact error for
every row. A rejected sampler is a recorded result, not a reason to silently
substitute another sampler.
## Pose transfer workflow
The [`repose/`](repose/) subdirectory contains a separate multi-reference
workflow for pose transfer. It keeps the character image in `Char073.png`,
extracts a DWPose image from `pose_reference.jpg`, and feeds the character
image plus pose map into FLUX.3 Klein. It is deliberately separate from the
one-reference sampler matrix because it changes the conditioning topology.
+41
View File
@@ -0,0 +1,41 @@
# FLUX.3 Klein DWPose repose
This workflow transfers a pose onto a character reference using the locally
installed FLUX.3 Klein model. It is not a native FLUX ControlNet workflow:
the seated/pose reference is first converted into a DWPose image, then the
character image and pose map are supplied as two FLUX.3 reference latents.
## Inputs
- `Char073.png` — character image expected in the ComfyUI input directory.
- `pose_reference.jpg` — included pose reference.
The character file is intentionally referenced by its existing server-side
name rather than copied into this repository. Replace node `1`'s filename if
using another character.
## Tested settings
- Model: `flux3-image-klein-int8-convrot.safetensors`
- Text encoder: `qwen3-vl-4b-heretic_int8.safetensors`
- VAE: `flux2-vae.safetensors`
- Sampler: Euler ancestral
- Steps: 16
- Guidance: 1.0
- Canvas: 1616×2464 (3.98 MP), portrait ratio
- DWPose: body and hands enabled, face disabled
The workflow is object-agnostic and uses the prompt to clarify which reference
controls identity/clothing and which controls pose. Results remain model-level
guidance rather than guaranteed skeletal constraints; hands, clothing and
background can still be reinterpreted.
## Wiring
`Char073.png` → image scale → VAE encode → ReferenceLatent 1
`pose_reference.jpg` → DWPose → VAE encode → ReferenceLatent 2
`ReferenceLatent 2` → FLUX.3 Klein sampler → VAE decode → Save Image
Use `workflow.json` with the ComfyUI API or recreate the same graph in the
UI. The included `results-char073-kneeling-4mp.png` is the accepted test
result for the included pose reference.
Binary file not shown.

After

Width:  |  Height:  |  Size: 23 KiB

@@ -0,0 +1,21 @@
{
"test_id": "F3K-REPOSE-CHAR073-KNEELING-4MP",
"status": "completed",
"server": "http://192.168.1.202:8188",
"model": "flux3-image-klein-int8-convrot.safetensors",
"text_encoder": "qwen3-vl-4b-heretic_int8.safetensors",
"vae": "flux2-vae.safetensors",
"character_image": "Char073.png",
"pose_image": "pose_reference.jpg",
"pose_preprocessor": "DWPreprocessor; body and hands enabled; face disabled",
"resolution": {"width": 1616, "height": 2464, "megapixels": 3.981824},
"steps": 16,
"guidance": 1.0,
"sampler": "euler_ancestral",
"seed": 731104,
"prompt_id": "5097d273-9143-4a68-b512-3470c3392554",
"output": "results-char073-kneeling-4mp.png",
"output_sha256": "cd5ff3bba77310ba817cba5f39c25cbca3282c51605734be896106e67f7ee127",
"pose_reference_sha256": "76cdf19e2439b1de25e354afa3d8a0a9c9dc8c9da91d7f72822f222fc565b37d",
"notes": "Successful DWPose-guided kneeling side-stretch transfer. Character file remains a server-side input dependency."
}
Binary file not shown.

After

Width:  |  Height:  |  Size: 5.8 MiB

@@ -0,0 +1,24 @@
{
"1": {"class_type":"LoadImage","inputs":{"image":"Char073.png"}},
"2": {"class_type":"LoadImage","inputs":{"image":"pose_reference.jpg"}},
"3": {"class_type":"ImageScaleToTotalPixels","inputs":{"image":["1",0],"upscale_method":"lanczos","megapixels":1.0,"resolution_steps":1}},
"4": {"class_type":"ImageScaleToTotalPixels","inputs":{"image":["2",0],"upscale_method":"lanczos","megapixels":1.0,"resolution_steps":1}},
"5": {"class_type":"DWPreprocessor","inputs":{"image":["4",0],"detect_hand":"enable","detect_body":"enable","detect_face":"disable","resolution":768,"bbox_detector":"yolox_l.onnx","pose_estimator":"dw-ll_ucoco_384.onnx","scale_stick_for_xinsr_cn":"disable"}},
"6": {"class_type":"UNETLoader","inputs":{"unet_name":"flux3-image-klein-int8-convrot.safetensors","weight_dtype":"default"}},
"7": {"class_type":"CLIPLoader","inputs":{"clip_name":"qwen3-vl-4b-heretic_int8.safetensors","type":"flux3","device":"default"}},
"8": {"class_type":"VAELoader","inputs":{"vae_name":"flux2-vae.safetensors"}},
"9": {"class_type":"VAEEncode","inputs":{"pixels":["3",0],"vae":["8",0]}},
"10": {"class_type":"VAEEncode","inputs":{"pixels":["5",0],"vae":["8",0]}},
"11": {"class_type":"CLIPTextEncode","inputs":{"text":"Create a photorealistic portrait of the exact adult woman in image 1, preserving her face, hair, skin tone, white outfit, accessories and proportions. Image 2 is a DWPose skeleton map and controls only the pose: one knee grounded, the other leg bent forward, torso leaning diagonally, one arm extended straight overhead and the other reaching down naturally. Do not copy clothing, identity, face, hair, background or colours from image 2. Preserve image 1 character and outfit. Keep one woman, anatomically correct hands and limbs, realistic weight and fabric folds, warm indoor photographic lighting, natural skin texture, shallow depth of field. No duplicate person, extra limbs, text or watermark.","clip":["7",0]}},
"12": {"class_type":"FluxGuidance","inputs":{"conditioning":["11",0],"guidance":1.0}},
"13": {"class_type":"ReferenceLatent","inputs":{"conditioning":["12",0],"latent":["9",0]}},
"14": {"class_type":"ReferenceLatent","inputs":{"conditioning":["13",0],"latent":["10",0]}},
"15": {"class_type":"BasicGuider","inputs":{"model":["6",0],"conditioning":["14",0]}},
"16": {"class_type":"KSamplerSelect","inputs":{"sampler_name":"euler_ancestral"}},
"17": {"class_type":"BasicScheduler","inputs":{"model":["6",0],"scheduler":"simple","steps":16,"denoise":1.0}},
"18": {"class_type":"RandomNoise","inputs":{"noise_seed":731104}},
"19": {"class_type":"EmptyLatentImage","inputs":{"width":1616,"height":2464,"batch_size":1}},
"20": {"class_type":"SamplerCustomAdvanced","inputs":{"noise":["18",0],"guider":["15",0],"sampler":["16",0],"sigmas":["17",0],"latent_image":["19",0]}},
"21": {"class_type":"VAEDecode","inputs":{"samples":["20",0],"vae":["8",0]}},
"22": {"class_type":"SaveImage","inputs":{"images":["21",0],"filename_prefix":"fluxExperiments/repose/FLUX3-repose-Char073-kneeling-4mp"}}
}
+1
View File
@@ -13,6 +13,7 @@ This directory contains the best tested workflows for the available hardware. Ch
- [`flux/`](I2I/flux/) — Flux.2 Klein photorealistic conversion.
- [`qwen/`](I2I/qwen/) — Qwen Image 2.1 photorealistic conversion.
- [`flux3klein/`](I2I/flux3klein/) — FLUX.3 Klein reference-image tests.
- [`repose/`](I2I/flux3klein/repose/) — FLUX.3 Klein + DWPose reference-pose workflow.
- [`I2IPlusRefiner/`](I2IPlusRefiner/) — edit-model output followed by a separate refiner, with base/refined image pairs.
- [`MultipleRef2I/`](MultipleRef2I/) — multi-reference image composition.
- [`flux/`](MultipleRef2I/flux/) — Flux.2 Klein best multi-reference workflow.