Qwen-Image 2.1 NSFW / Pose Pipeline
Local photorealistic NSFW + pose control: ComfyUI + Qwen-Image 2.1 (int8 convrot)
- NSFW LoRAs + a 136-pose VNCCS 3D pose library.
Path convention: everything in this skill is relative to the skill root
(this directory). Scripts in scripts/, workflow graphs in workflows/,
author prompts in prompts/, reference images in references/,
author sample data (originals + metadata) in downloads/samples/.
ComfyUI-side locations (models, LoRAs, pose library) are configurable via env
vars — see install/setup.md and §2. The default host for ComfyUI is
http://127.0.0.1:8188 (override with COMFYUI_URL).
This file is self-contained: setup, recipe, pose selection, and a first-run
author case are all inside.
0. First run — follow the author's case (recommended)
The fastest way to trust the pipeline is re-running the author's own sample
prompt and comparing against the author's published image (verified 2026-10-12):
python3 scripts/qwen_pose_test.py \
--mode t2i --name author_case_shower \
--prompt "$(cat prompts/taws_v13_shower.txt)" \
--seed 80964 \
--steps 50 --cfg 2.5 --sampler er_sde --scheduler beta \
--width 1152 --height 1728 \
--lora "qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors" \
--lora-strength 1.0 \
--out outputs/author_case/shower.png
- Other cases:
prompts/taws_v13_massage.txt (30 steps, same seed, per its
.params.json) and prompts/thesealpacas_v2_selfie.txt (CFG 4.0 +
thesealpacas-nsfw-v2 @0.8; its .txt ends with a [negative] section —
pass that to --negative).
- Author originals to compare against:
downloads/samples/*_author.jpg
(shower / massage / selfie). Full author metadata (prompts/params/seed/
complete ComfyUI graph) in downloads/samples/phase5_meta.json.
1. Routing
| Request |
Route |
Tool |
| New sexy/NSFW image from a text idea |
① T2I + NSFW LoRA (mandatory) |
scripts/qwen_pose_test.py --mode t2i |
| Change pose of an existing character image (keep face/clothes) |
② Edit |
scripts/qwen_pose_test.py --mode edit |
| Change pose from a 3D mannequin render (script, no browser) |
③ Pose transfer |
scripts/run_pose_transfer.py |
| Exact pose from the 3D pose library (136 poses) |
④ VNCCS PoseStudio |
scripts/run_vnccs_ui.py (browser automation) |
| Do NOT use: a plain photo or illustration as pose reference — the model composites it as a second person, or redraws the reference person entirely (tested twice); use a textureless 3D mannequin render with route ③/④ instead |
|
|
Outputs land under ./outputs/ (the scripts create directories as needed).
2. Setup — what this package contains and what to install
2.1 In this package
| Path |
Content |
SKILL.md |
this file |
scripts/ |
doctor.py (read-only environment self-check: ComfyUI, models, LoRAs, VNCCS, pose library, browser stack — every check degrades to WARN/SKIP with a fix hint), qwen_pose_test.py (routes ①②), run_pose_transfer.py (route ③: pose transfer from a 3D mannequin render, no browser), run_vnccs_ui.py (route ④), make_custom_pose.py (new pose from a library pose by bone rotation), run_phase5.py + build_phase5_sheet.py (case re-run & comparison sheet), media_lock.py (shared media lock, bundled), browser_fallback.py (standalone CDP fallback when the OpenClaw adapter is absent), validate_package.py (offline package integrity check) |
workflows/ |
qwen21_t2i.json, qwen21_edit.json (official Qwen-Image 2.1 graphs), VNCCS_PoseStudio_QI2.1.json |
prompts/ |
the 3 author prompts + .params.json (the exact local-test settings) |
references/ |
pose_examples/ (L0/L1/L2 seaside examples), src_standing.png (default character for route ④) |
downloads/samples/ |
author originals *_author.jpg (shower/massage/selfie) + phase5_meta.json (prompt, params, seed, complete graph JSONs) + the 3 raw ComfyUI graph files — used by scripts/run_phase5.py & build_phase5_sheet.py |
pose_library/local_user_poses/Uncategorized/ |
55 curated extra pose JSONs + previews — copy into ComfyUI's PoseLibrary/local_user_poses/Uncategorized/ after installing the node |
install/setup.md |
full install guide (models, LoRAs, node, pose repos, commands) |
2.2 ComfyUI-side dependencies (see install/setup.md for exact commands)
- Base models (3 files) from HF
Comfy-Org/Qwen-Image-2.1:
diffusion_models/qwen_image_2.1_int8_convrot.safetensors,
text_encoders/qwen3vl_8b_int8_convrot.safetensors,
vae/qwen_image_2.1_vae_bf16.safetensors — into the ComfyUI model dirs.
- LoRAs (into ComfyUI
models/loras/):
| File (size) |
Use |
Source |
qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors (608M) |
default: anatomy + pose, sharpest |
Civitai model 2986844 — tawsLoRa "Qwen Image 2.1 - FemaledVaginus-NSFW" |
qwen2_1-femaledVaginus-nsfw-v1.3-fine.safetensors (608M) |
same family, v1.3 |
Civitai model 2986844 |
thesealpacas-nsfw-v2.safetensors (79.7M) |
most reveal, "influencer" face drift |
Civitai model 2958918 |
nsfw-posture-v3.safetensors (84M) |
strongest pose/body emphasis |
Civitai model 2988215 |
qwen21-nsfw-image-edit.safetensors (84M) |
edit-route undress/pose changes |
HF RunningHubAI/Qwen-Image-2.1-NSFW-Image-Edit (mirror of oreolove "Nude Edit"; original Civitai page delisted) |
VNCCS_QI2_PoseStudioV1.1.safetensors (159M) |
routes ③ and ④ (file name selectable via the VNCCS_LORA env var) |
HF MIUProject/VNCCS_PoseStudio_QI2.1 (same repo as the workflow JSON) |
- VNCCS_Utils custom node: https://github.com/AHEKOT/ComfyUI_VNCCS_Utils →
clone into
ComfyUI/custom_nodes/, then restart ComfyUI once.
- Pose library: 4 HuggingFace repos (the public catalog as of 2026-10-12):
Totemistyk/General_Poses_PoseStudio (50), Totemistyk/Ballet_PoseStudio
(3 in manifest; the repo actually ships 58 — register + refresh, then copy
this package's 55 curated extras into local_user_poses/Uncategorized/),
MIUProject/VNCCS_PoseLibrary_Main (16), MIUProject/MySuperPoseRepo (12).
Register: curl -X POST 127.0.0.1:8188/vnccs/pose_library/repositories/add -d '{"repo_id":"<owner>/<repo>"}'. Total after setup: 136.
A repo refresh deletes local files not in the remote manifest — extra/
curated poses belong in PoseLibrary/local_user_poses/<Category>/
(always listed, never auto-cleaned). Pose list API: GET /vnccs/pose_library/list.
- Route ④ extras: any CDP Chrome (default
127.0.0.1:18800) +
browser_client.py + playwright. On the original host these live in
~/.openclaw/tools/browser/; elsewhere point run_vnccs_ui.py at an
equivalent CDP endpoint (it only needs connect_sync(playwright) to return a
browser with one context — a minimal stand-in is fine).
- media_lock: bundled in
scripts/; both scripts import it first from the
package, then fall back to ~/.openclaw/tools.
The base model is qwen_image_2.1_int8_convrot.safetensors — the same model
the Civitai authors of these LoRAs used, so their published prompts/params
transfer well. (Note: one of the three author samples actually ran on a
different base with 5 stacked LoRAs; we reproduced it with the single
thesealpacas-v2 LoRA on the convrot base — see Parameter provenance below.)
Download tips (verified): Civitai via the official CLI or
curl -H "Authorization: Bearer <token>" <file-url>; HuggingFace via
hf-mirror.com directly (no proxy — a proxy causes SSL_ERROR).
3. Route ① — T2I (the "author recipe", validated 2026-10-12)
3.1 Mandatory recipe — do not weaken silently
This recipe reproduces Civitai showcase-level results. All of the following are
part of the recipe; a deviation changes the result and must be told to the user:
- Sampler
er_sde + scheduler beta (not euler/simple)
- CFG 2.5 (the A1 reference; 2.5–4 works, CFG 1 starves the LoRA)
- Steps = the reference sample's value: 50 (A1 provenance, table below)
- Size 1152×1728 portrait 2:3 (
--width 1152 --height 1728); keep
--resolution 1024 (that flag feeds TextEncodeQwenImage21, it is not pixel size)
- A LoRA is always on for route ①. Default:
qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors @ 1.0
- The prompt is the main lever — it must follow the author structure
(§3.2) at the author's density (300–1500 characters, explicit terms).
Short/keyword prompts and SFW-generic phrasing are what make output look bland.
- Pass the reference sample's seed when you claim "reproduction"
(A1 seed = 80964). The harness default seed (424242) is fine for new
creative runs but must not be labelled as the author's result.
Parameter provenance (source: scripts/run_phase5.py + each run's *.png.json):
| Run record |
Steps |
CFG |
Seed |
LoRA strength (local test) |
| A1 shower (primary reference) |
50 |
2.5 |
80964 |
1.0 |
| A2 massage (separate sample) |
30 |
2.5 |
80964 |
1.0 |
| B1 selfie |
50 |
4.0 |
124880844757561 |
0.8 (thesealpacas-v2) |
| C1 control |
50 |
2.5 |
80964 |
no LoRA |
Preserve each sample's settings as a unit; do not mix one sample's prompt with
another sample's params and present it as "the author recipe".
3.2 Prompt structure (copy the pattern, fill the scene)
The working prompts all contain, in this order — write all 7, no omissions:
- Camera: "A hyper-realistic photograph / low-angle shot / top-down overhead shot", lens and angle
- Subject: age, build, hair (color, wet/dry, placement)
- Pose: pick one from the pose ladder (§4) and spell out each limb (which leg bent, where the hands are, what the head/gaze does)
- Setting + lighting: "warm steamy light", "neon bokeh", "soft morning light through sheer curtains"
- Skin: "glistening", "visible pores", "natural skin texture", "slick oily sheen"
- Anatomy/reveal where desired — the LoRAs were trained on explicit terms
(e.g. "labia majora ... innie distinctly visible"); generic words ("seductive")
do almost nothing
- Finish: "8k resolution, cinematic lighting, candid 'caught in the moment'
atmosphere, photorealistic anatomy"
Template (a real author prompt runs 300–1500 chars):
A hyper-realistic, {camera} shot {subject with age/build/hair state},
{one pose from §4, limb by limb}. {Setting + lighting}. Her skin is
{glistening/visible pores/slick oily sheen}. {explicit anatomy/reveal wording}.
{face/expression}. 8k resolution, cinematic lighting, candid "caught in the
moment" atmosphere, photorealistic anatomy.
Before writing, load a verified author prompt and rewrite it for your scene —
prompts/*.txt (+ matching .params.json). Writing from scratch is what makes
results look generic.
3.3 Deviations — tell, don't silently do it
If the user asks for a shorter prompt, a SFW-generic prompt, "no LoRA", fewer
steps, or a different sampler — and if you (the agent) ever need to deviate
from the recipe for any reason (model refuses a token, out-of-memory, etc.):
say the expected cost out loud, show the evidence, and proceed only after
confirmation. Never present a weakened run as "the recipe's result"; label what
was changed next to the delivered image. Evidence to cite:
- True controlled experiment (single variable = LoRA):
C1 (author prompt,
50 steps, seed 80964, no LoRA) vs A1 (identical prompt/params + LoRA):
no-LoRA still looks showcase-grade as a photograph, but the LoRA is what adds
the skin/gloss/anatomy detail the author's images have. So "no LoRA" ≈
"good photograph, not the author's texture".
- Visual demo (multiple variables: prompt + outfit + LoRA + seed):
references/pose_examples/01_seaside.png (L0, no LoRA) vs
01b_seaside_lora.png (L1 + LoRA) vs 06_l2_squat_lowangle.png
(L2 + low angle + LoRA) — shows the ladder of how much the full recipe adds.
It demonstrates the effect but does not isolate any single factor.
3.4 Known limits of the harness
- One LoRA per run.
qwen_pose_test.py injects a single LoraLoaderModelOnly
(node id 99) between UNETLoader and KSampler. Civitai showcase workflows
often stack 3–5 LoRAs; to reproduce those, chain extra LoraLoaderModelOnly
nodes in build_workflow().
- No refine pass / no EasyCache. Authors add a 2× lanczos upscale + a 25-step
denoise 0.35 pass and EasyCache; QwenImage21Cache is broken on the
original host (aimdo memory compile error) and the route ④ runner removes it.
A/B runs matched the authors' look without the refine pass — add it only when
a job needs extra sharpness.
4. Pose selection (the "boldness scale") — match the user's request
Read the user's request (clothes state, desired mood, scene) → pick a tier →
give a concrete pose, not a vague one. "Bold" is three levers combined:
pose openness × camera angle × clothing/wet state.
4.1 Tiers (English wording to use in route ① prompts)
- L0 closed / editorial (clothed SFW): standing with weight on one leg, hand
on hip or in pocket; sitting on a chair, legs crossed at the ankles; leaning on
a railing/wall, looking back over the shoulder; mid-stride walking.
- L1 open (swimwear / wet / relaxed): crouching with knees apart, hands on
the floor; sitting on the floor with knees opened to one side; lying on her
side in bed, propped on an elbow, looking back over the shoulder; kneeling with
hands behind; reclining on a sofa with legs open.
- L2 bold (author style — always pair with the NSFW LoRA + low angle):
standing with legs apart, shot from a low angle looking up; lying prone with
hips lifted toward the camera; sitting with legs wide open, leaning forward on
the elbows; kneeling on one knee with the other leg extended to the side, torso
arched back; lying on her back with legs open, one hand behind the head.
- L3 extreme / cropped: extreme low angle looking up between open legs;
overhead shot lying flat with legs open; cropped framing (waist-up or
mid-thigh) to emphasize the pose.
4.2 Suggestion table (user request → tier → concrete pose)
| User asks for (examples) |
Tier |
Route ④ library pose families (--list-poses names) |
Route ① prompt wording |
| 全着装、自然/编辑感 (standing, portrait, office/street) |
L0 |
Standing / Sitting / Walking / Close Up |
standing with weight on one leg, hand in pocket; seated on a chair, legs crossed at the ankles |
| 泳装/湿身/海滩/浴室 (swimwear, wet, relaxed) |
L1 |
Squatting / Kneeling / Laying Back |
crouching with knees apart, hands on the floor, looking back over the shoulder |
| 大胆/NSFW (nude + LoRA, low angle, "author style") |
L2 |
Kneeling / Laying Back / Prone / Splits / Crawling / Dancing on knees |
standing with legs apart from a low angle; prone with hips lifted toward the camera; seated wide, leaning forward on elbows |
| 极端/特写构图 (crop, extreme low, overhead) |
L3 |
Splits / any L2 family + crop framing |
extreme low angle looking up between open legs; overhead, lying flat with legs open |
Rules:
- Decide the reveal first: clothed → L0; swimwear/wet → L1; nude + NSFW LoRA → L2.
Pairing the LoRA with an L0 pose is the most common cause of "bland" NSFW
output (observed, not a proven law).
- Camera angle amplifies the pose (empirical heuristic): low angle ≈ +1
boldness, overhead ≈ +1. "Low angle + L1" ≈ "eye level + L2" — this is why
the author's shower/massage shots read as bold despite simple poses.
- One pose per prompt, limb by limb; two competing pose descriptions get blended.
- When the user names a pose they saw somewhere, search the library first:
python3 scripts/run_vnccs_ui.py --list-poses (136 entries with file paths).
5. Route ② — Edit (pose change with identity preservation)
python3 scripts/qwen_pose_test.py \
--mode edit --name pose_change --input /absolute/base.png \
--prompt "change her pose to kneeling on one knee, hands on hips, keep face hair and outfit" \
--cfg 3 --sampler er_sde --scheduler beta --steps 20 \
--out outputs/pose_change.png
CFG 3 is more stable for edit mode. SFW instruction prompts keep NSFW LoRAs from
triggering; add --lora qwen21-nsfw-image-edit.safetensors only when undressing.
6. Route ③ — Scripted pose transfer (3D mannequin reference)
python3 scripts/run_pose_transfer.py \
--char /absolute/character.png \
--pose-ref /absolute/doll_render.png \
--out outputs/pose_transfer.png [--seed N] [--steps 25]
The pose source must be a textureless 3D mannequin render — e.g. the
doll_*.png captures that route ④ saves under the ComfyUI output dir, or any
3D poser render of the target pose. Photos and illustrations as pose references
do NOT work: the model composites the reference person as a second person, or
redraws the reference person entirely (retested 2026-10-16 — all variants
failed).
Verified recipe (2026-10-16) — the script sets it up automatically:
- dual-image edit with image 1 = mannequin (pose) and image 2 = character (identity);
- prompt
replace the pose of <image 2> with the pose of <image 1>. keep the character of <image 2>;
- VNCCS PoseStudio LoRA @ 1.0 is mandatory — without it the mannequin is
ignored entirely (3 no-LoRA controls);
- 25 steps, cfg 1, euler/simple, 1024 px. Single person, identity kept, pose
follows the render; stable across seeds (sit ×3 seeds + crawling /
kneel-side / lean-back all passed). Camera framing can differ slightly from
the render; producing new mannequin renders still requires a route ④
browser session — route ③ reuses captures you already have.
7. Route ④ — VNCCS PoseStudio (exact pose from the 3D library)
The VNCCS_PoseStudio node needs the ComfyUI web frontend for its 3D "live
capture"; the headless API fails with "Live capture timed out". So the runner
drives a CDP browser: load the official graph → set character image + pose
JSON → queue → collect outputs from disk. ~60–90 s each.
# list all 136 poses with resolved file paths
python3 scripts/run_vnccs_ui.py --list-poses
# run one pose on a character image (+ optional scene prompt)
python3 scripts/run_vnccs_ui.py --pose "squatting leaning back" \
[--char /absolute/face.png] [--prompt "on a rooftop at night, city bokeh"] [--timeout 900]
--pose takes the name/slug case-insensitively (crawling, Squatting Leaning Back, arabesque). 9 names exist in more than one repository
(Standing A, CloseUp A, …); qualify with <repo>/<category>/<name>
(exact keys printed by --list-poses, e.g.
miuproject_vnccs_poselibrary_main/standing/standing_a).
Scene prompt (--prompt): the capture prompt is rebuilt at execution
time from the scene document's export block (default template
Draw character from image2\n<lighting>\n<user_prompt>); --prompt "..."
injects export.user_prompt before the graph loads. Editing the
TextEncodeQwenImage21 widget or adding export.user_prompt to a library
pose file has no effect. To change the scene of an existing route ④ result,
re-run it through route ② (qwen_pose_test.py --mode edit).
The runner reaches a CDP browser through the OpenClaw adapter when present,
otherwise through the bundled scripts/browser_fallback.py (reuses a
reachable COMFY_BROWSER_CDP endpoint, else launches a throwaway headless
Chromium; COMFY_BROWSER_MODE=auto|attach|launch).
The runner re-indexes the pose library on every call — newly downloaded poses
are usable immediately. Each run writes <pose>_<stamp>.meta.json next to
the outputs and never clobbers previous results (unique comfy prefix per run).
Output: ./outputs/vnccs_ui/ (override with VNCCS_UI_OUT). The 3D doll is a
bare MakeHuman mannequin — nude prompts + pose transfer can produce nude
output (expected: the model reads the doll's state).
Custom poses from code: scripts/make_custom_pose.py loads a library
pose into the PoseStudio editor, adds bone rotations, and saves through the
editor's own Save (JSON + preview in
PoseLibrary/local_user_poses/poses/<category>/). The new pose works with
run_vnccs_ui.py --pose at once. --list-bones prints the base pose's
euler angles; pass the full library id when a name is not unique.
python3 scripts/make_custom_pose.py \
--base "MIUProject__VNCCS_PoseLibrary_Main/poses/Standing/Standing A" \
--name "Arms Out" --rotate upperarm_l:1:90 --rotate upperarm_r:1:-90
Pose fidelity caveat: each pose JSON carries its own cameraParams
(yaw/pitch/zoom); the generated image inherits that viewpoint. Poses whose
shape only reads from another angle (e.g. a true arabesque's extended back
leg) come out front-facing and ambiguous — rotate the doll in the UI and
re-run, or pick a pose that reads from the front.
8. Environment & Safety
- ComfyUI as a service (original host:
systemctl --user status|restart comfyui.service). After installing new custom nodes, one restart is required.
- The route scripts take the bundled media lock in shared mode
(
media_lock(exclusive=False, wait_timeout=...)) so they don't fight video
jobs on a shared machine; route ④ wraps the whole browser run in it.
- On the original host, video tasks hold the lock exclusively (~06:45 / 09:00 /
11:30 openclaw windows); expect waits, not conflicts.
- If a graph includes
QwenImage21Cache and the host errors on it
(aimdo memory compile error), remove the node and rewire LoRA → KSampler
(the route ④ runner does this automatically in the browser).
- Never modify the workflow JSONs in this package's
workflows/ while a run is
in flight; the harness works on in-memory copies anyway.
- A 50-step 1152×1728 run takes ~200–400 s on the int8 model; a 30-step edit
run ~60–120 s.
9. Verification
- Check the harness printed
ok: true and the output file is non-empty.
- Visually inspect with an image-reading tool (composition, pose match,
anatomy, face). For pose-transfer runs, compare against the reference.
- For reproductions of published images, compare side-by-side with the author
original (
downloads/samples/*_author.jpg; comparison-sheet
example: scripts/build_phase5_sheet.py).
- Record prompt/seed/steps/LoRA in the run's
*.json meta file (harnesses
write one automatically for routes ①②③④).
- When a recipe deviation was made (weaker prompt, no LoRA, fewer steps,
default seed), label the output and state the deviation next to it in the
delivery message.
Last full verification: 2026-10-12 — route ① (4 author-prompt reproductions),
route ② (9/9 pose edits with identity preserved), route ④ (6 pose runs:
crawling, kneeling side, leaning back, sitting side, squatting leaning back,
arabesque — the last one verified for output only, pose-shape match unverified;
its doll capture turned out to be a stale legacy capture),
pose index count 136 == GET /vnccs/pose_library/list, pose-ladder A/B images
in references/pose_examples/.
Follow-up checks 2026-10-14: route ④ --prompt (scene follows the prompt, pose
preserved), bundled browser fallback launch path, and custom pose creation via
scripts/make_custom_pose.py (save + output verified).
Follow-up checks 2026-10-16 (route ③): 6 mannequin-reference transfers passed
(sit ×3 seeds, crawling, kneel-side, lean-back): single person, identity kept,
pose follows the render; 3 no-LoRA controls ignored the mannequin; photo and
illustration references still fail (composite / redraw of the reference
person). Script scripts/run_pose_transfer.py added and end-to-end tested.
1---2name: arekejoker-qwen21-nsfw-pose-skill3description: "Use when generating sexy/NSFW or posed images with local Qwen-Image 2.1 — pose/NSFW LoRAs (tawsLoRa femaledVaginus, TheseAlpacas, posture-v3), pose transfer from an existing image, or exact pose control via VNCCS PoseStudio 3D pose library. Triggers: 性感姿势, 摆姿势, NSFW 出图, 换姿势, 姿势控制, pose transfer, pose lora, VNCCS, 人偶姿势, 姿势库. Independent of any general image-generation skill (which handles SFW text/edit/cutout work)."4license: Apache-2.05metadata:6 source: https://skillmd.com/skills/gabrielmoreira/arekejoker-qwen21-nsfw-pose-skill7---89# Qwen-Image 2.1 NSFW / Pose Pipeline1011Local photorealistic NSFW + pose control: ComfyUI + Qwen-Image 2.1 (int8 convrot)12+ NSFW LoRAs + a 136-pose VNCCS 3D pose library.1314**Path convention:** everything in this skill is relative to the skill root15(this directory). Scripts in `scripts/`, workflow graphs in `workflows/`,16author prompts in `prompts/`, reference images in `references/`,17author sample data (originals + metadata) in `downloads/samples/`.18ComfyUI-side locations (models, LoRAs, pose library) are configurable via env19vars — see `install/setup.md` and §2. The default host for ComfyUI is20`http://127.0.0.1:8188` (override with `COMFYUI_URL`).2122This file is self-contained: setup, recipe, pose selection, and a first-run23author case are all inside.2425## 0. First run — follow the author's case (recommended)2627The fastest way to trust the pipeline is re-running the author's own sample28prompt and comparing against the author's published image (verified 2026-10-12):2930```bash31python3 scripts/qwen_pose_test.py \32 --mode t2i --name author_case_shower \33 --prompt "$(cat prompts/taws_v13_shower.txt)" \34 --seed 80964 \35 --steps 50 --cfg 2.5 --sampler er_sde --scheduler beta \36 --width 1152 --height 1728 \37 --lora "qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors" \38 --lora-strength 1.0 \39 --out outputs/author_case/shower.png40```4142- Other cases: `prompts/taws_v13_massage.txt` (30 steps, same seed, per its43 `.params.json`) and `prompts/thesealpacas_v2_selfie.txt` (CFG 4.0 +44 `thesealpacas-nsfw-v2` @0.8; its `.txt` ends with a `[negative]` section —45 pass that to `--negative`).46- Author originals to compare against: `downloads/samples/*_author.jpg`47 (shower / massage / selfie). Full author metadata (prompts/params/seed/48 complete ComfyUI graph) in `downloads/samples/phase5_meta.json`.4950## 1. Routing5152| Request | Route | Tool |53| --- | --- | --- |54| New sexy/NSFW image from a text idea | ① T2I + NSFW LoRA (mandatory) | `scripts/qwen_pose_test.py --mode t2i` |55| Change pose of an existing character image (keep face/clothes) | ② Edit | `scripts/qwen_pose_test.py --mode edit` |56| Change pose from a 3D mannequin render (script, no browser) | ③ Pose transfer | `scripts/run_pose_transfer.py` |57| Exact pose from the 3D pose library (136 poses) | ④ VNCCS PoseStudio | `scripts/run_vnccs_ui.py` (browser automation) |58| Do NOT use: a plain photo or illustration as pose reference — the model composites it as a second person, or redraws the reference person entirely (tested twice); use a textureless 3D mannequin render with route ③/④ instead | | |5960Outputs land under `./outputs/` (the scripts create directories as needed).6162## 2. Setup — what this package contains and what to install6364### 2.1 In this package6566| Path | Content |67| --- | --- |68| `SKILL.md` | this file |69| `scripts/` | `doctor.py` (read-only environment self-check: ComfyUI, models, LoRAs, VNCCS, pose library, browser stack — every check degrades to WARN/SKIP with a fix hint), `qwen_pose_test.py` (routes ①②), `run_pose_transfer.py` (route ③: pose transfer from a 3D mannequin render, no browser), `run_vnccs_ui.py` (route ④), `make_custom_pose.py` (new pose from a library pose by bone rotation), `run_phase5.py` + `build_phase5_sheet.py` (case re-run & comparison sheet), `media_lock.py` (shared media lock, bundled), `browser_fallback.py` (standalone CDP fallback when the OpenClaw adapter is absent), `validate_package.py` (offline package integrity check) |70| `workflows/` | `qwen21_t2i.json`, `qwen21_edit.json` (official Qwen-Image 2.1 graphs), `VNCCS_PoseStudio_QI2.1.json` |71| `prompts/` | the 3 author prompts + `.params.json` (the exact local-test settings) |72| `references/` | `pose_examples/` (L0/L1/L2 seaside examples), `src_standing.png` (default character for route ④) |73| `downloads/samples/` | author originals `*_author.jpg` (shower/massage/selfie) + `phase5_meta.json` (prompt, params, seed, complete graph JSONs) + the 3 raw ComfyUI graph files — used by `scripts/run_phase5.py` & `build_phase5_sheet.py` |74| `pose_library/local_user_poses/Uncategorized/` | 55 curated extra pose JSONs + previews — copy into ComfyUI's `PoseLibrary/local_user_poses/Uncategorized/` after installing the node |75| `install/setup.md` | full install guide (models, LoRAs, node, pose repos, commands) |7677### 2.2 ComfyUI-side dependencies (see `install/setup.md` for exact commands)7879- **Base models (3 files)** from HF `Comfy-Org/Qwen-Image-2.1`:80 `diffusion_models/qwen_image_2.1_int8_convrot.safetensors`,81 `text_encoders/qwen3vl_8b_int8_convrot.safetensors`,82 `vae/qwen_image_2.1_vae_bf16.safetensors` — into the ComfyUI model dirs.83- **LoRAs** (into ComfyUI `models/loras/`):8485| File (size) | Use | Source |86| --- | --- | --- |87| `qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors` (608M) | default: anatomy + pose, sharpest | Civitai model 2986844 — tawsLoRa "Qwen Image 2.1 - FemaledVaginus-NSFW" |88| `qwen2_1-femaledVaginus-nsfw-v1.3-fine.safetensors` (608M) | same family, v1.3 | Civitai model 2986844 |89| `thesealpacas-nsfw-v2.safetensors` (79.7M) | most reveal, "influencer" face drift | Civitai model 2958918 |90| `nsfw-posture-v3.safetensors` (84M) | strongest pose/body emphasis | Civitai model 2988215 |91| `qwen21-nsfw-image-edit.safetensors` (84M) | edit-route undress/pose changes | HF `RunningHubAI/Qwen-Image-2.1-NSFW-Image-Edit` (mirror of oreolove "Nude Edit"; original Civitai page delisted) |92| `VNCCS_QI2_PoseStudioV1.1.safetensors` (159M) | routes ③ and ④ (file name selectable via the `VNCCS_LORA` env var) | HF `MIUProject/VNCCS_PoseStudio_QI2.1` (same repo as the workflow JSON) |9394- **VNCCS_Utils custom node**: https://github.com/AHEKOT/ComfyUI_VNCCS_Utils →95 clone into `ComfyUI/custom_nodes/`, then restart ComfyUI once.96- **Pose library**: 4 HuggingFace repos (the public catalog as of 2026-10-12):97 `Totemistyk/General_Poses_PoseStudio` (50), `Totemistyk/Ballet_PoseStudio`98 (3 in manifest; the repo actually ships 58 — register + refresh, then copy99 this package's 55 curated extras into `local_user_poses/Uncategorized/`),100 `MIUProject/VNCCS_PoseLibrary_Main` (16), `MIUProject/MySuperPoseRepo` (12).101 Register: `curl -X POST 127.0.0.1:8188/vnccs/pose_library/repositories/add102 -d '{"repo_id":"<owner>/<repo>"}'`. Total after setup: 136.103 **A repo refresh deletes local files not in the remote manifest** — extra/104 curated poses belong in `PoseLibrary/local_user_poses/<Category>/`105 (always listed, never auto-cleaned). Pose list API: `GET /vnccs/pose_library/list`.106- **Route ④ extras**: any CDP Chrome (default `127.0.0.1:18800`) +107 `browser_client.py` + playwright. On the original host these live in108 `~/.openclaw/tools/browser/`; elsewhere point `run_vnccs_ui.py` at an109 equivalent CDP endpoint (it only needs `connect_sync(playwright)` to return a110 browser with one context — a minimal stand-in is fine).111- **media_lock**: bundled in `scripts/`; both scripts import it first from the112 package, then fall back to `~/.openclaw/tools`.113114The base model is `qwen_image_2.1_int8_convrot.safetensors` — the same model115the Civitai authors of these LoRAs used, so their published prompts/params116transfer well. (Note: one of the three author samples actually ran on a117different base with 5 stacked LoRAs; we reproduced it with the single118thesealpacas-v2 LoRA on the convrot base — see Parameter provenance below.)119120Download tips (verified): Civitai via the official CLI or121`curl -H "Authorization: Bearer <token>" <file-url>`; HuggingFace via122`hf-mirror.com` directly (no proxy — a proxy causes SSL_ERROR).123124## 3. Route ① — T2I (the "author recipe", validated 2026-10-12)125126### 3.1 Mandatory recipe — do not weaken silently127128This recipe reproduces Civitai showcase-level results. **All of the following are129part of the recipe; a deviation changes the result and must be told to the user:**130131- **Sampler** `er_sde` + scheduler `beta` (not euler/simple)132- **CFG 2.5** (the A1 reference; 2.5–4 works, CFG 1 starves the LoRA)133- **Steps = the reference sample's value: 50** (A1 provenance, table below)134- **Size 1152×1728 portrait 2:3** (`--width 1152 --height 1728`); keep135 `--resolution 1024` (that flag feeds `TextEncodeQwenImage21`, it is not pixel size)136- **A LoRA is always on for route ①.** Default:137 `qwen2_1-femaledVaginus-nsfw-v1_3-fine_L22-31.safetensors` @ **1.0**138- **The prompt is the main lever** — it must follow the author structure139 (§3.2) at the author's density (300–1500 characters, explicit terms).140 Short/keyword prompts and SFW-generic phrasing are what make output look bland.141- **Pass the reference sample's seed** when you claim "reproduction"142 (A1 seed = 80964). The harness default seed (424242) is fine for new143 creative runs but must not be labelled as the author's result.144145Parameter provenance (source: `scripts/run_phase5.py` + each run's `*.png.json`):146147| Run record | Steps | CFG | Seed | LoRA strength (local test) |148| --- | --- | --- | --- | --- |149| A1 shower (primary reference) | 50 | 2.5 | 80964 | 1.0 |150| A2 massage (separate sample) | 30 | 2.5 | 80964 | 1.0 |151| B1 selfie | 50 | 4.0 | 124880844757561 | 0.8 (thesealpacas-v2) |152| C1 control | 50 | 2.5 | 80964 | no LoRA |153154Preserve each sample's settings as a unit; do not mix one sample's prompt with155another sample's params and present it as "the author recipe".156157### 3.2 Prompt structure (copy the pattern, fill the scene)158159The working prompts all contain, in this order — write all 7, no omissions:1601. Camera: "A hyper-realistic photograph / low-angle shot / top-down overhead shot", lens and angle1612. Subject: age, build, hair (color, wet/dry, placement)1623. Pose: pick one from the pose ladder (§4) and spell out each limb (which leg bent, where the hands are, what the head/gaze does)1634. Setting + lighting: "warm steamy light", "neon bokeh", "soft morning light through sheer curtains"1645. Skin: "glistening", "visible pores", "natural skin texture", "slick oily sheen"1656. Anatomy/reveal where desired — the LoRAs were trained on explicit terms166 (e.g. "labia majora ... innie distinctly visible"); generic words ("seductive")167 do almost nothing1687. Finish: "8k resolution, cinematic lighting, candid 'caught in the moment'169 atmosphere, photorealistic anatomy"170171Template (a real author prompt runs 300–1500 chars):172173```174A hyper-realistic, {camera} shot {subject with age/build/hair state},175{one pose from §4, limb by limb}. {Setting + lighting}. Her skin is176{glistening/visible pores/slick oily sheen}. {explicit anatomy/reveal wording}.177{face/expression}. 8k resolution, cinematic lighting, candid "caught in the178moment" atmosphere, photorealistic anatomy.179```180181**Before writing, load a verified author prompt and rewrite it for your scene** —182`prompts/*.txt` (+ matching `.params.json`). Writing from scratch is what makes183results look generic.184185### 3.3 Deviations — tell, don't silently do it186187If the user asks for a shorter prompt, a SFW-generic prompt, "no LoRA", fewer188steps, or a different sampler — **and** if you (the agent) ever need to deviate189from the recipe for any reason (model refuses a token, out-of-memory, etc.):190say the expected cost out loud, show the evidence, and proceed only after191confirmation. Never present a weakened run as "the recipe's result"; label what192was changed next to the delivered image. Evidence to cite:193194- **True controlled experiment (single variable = LoRA):** `C1` (author prompt,195 50 steps, seed 80964, **no LoRA**) vs `A1` (identical prompt/params + LoRA):196 no-LoRA still looks showcase-grade as a photograph, but the LoRA is what adds197 the skin/gloss/anatomy detail the author's images have. So "no LoRA" ≈198 "good photograph, not the author's texture".199- **Visual demo (multiple variables: prompt + outfit + LoRA + seed):**200 `references/pose_examples/01_seaside.png` (L0, no LoRA) vs201 `01b_seaside_lora.png` (L1 + LoRA) vs `06_l2_squat_lowangle.png`202 (L2 + low angle + LoRA) — shows the ladder of how much the full recipe adds.203 It demonstrates the effect but does **not** isolate any single factor.204205### 3.4 Known limits of the harness206207- **One LoRA per run.** `qwen_pose_test.py` injects a single `LoraLoaderModelOnly`208 (node id 99) between `UNETLoader` and `KSampler`. Civitai showcase workflows209 often stack 3–5 LoRAs; to reproduce those, chain extra `LoraLoaderModelOnly`210 nodes in `build_workflow()`.211- **No refine pass / no EasyCache.** Authors add a 2× lanczos upscale + a 25-step212 `denoise 0.35` pass and `EasyCache`; `QwenImage21Cache` is broken on the213 original host (`aimdo memory compile error`) and the route ④ runner removes it.214 A/B runs matched the authors' look without the refine pass — add it only when215 a job needs extra sharpness.216217## 4. Pose selection (the "boldness scale") — match the user's request218219Read the user's request (clothes state, desired mood, scene) → pick a tier →220give a **concrete pose**, not a vague one. "Bold" is three levers combined:221**pose openness × camera angle × clothing/wet state**.222223### 4.1 Tiers (English wording to use in route ① prompts)224225- **L0 closed / editorial** (clothed SFW): standing with weight on one leg, hand226 on hip or in pocket; sitting on a chair, legs crossed at the ankles; leaning on227 a railing/wall, looking back over the shoulder; mid-stride walking.228- **L1 open** (swimwear / wet / relaxed): crouching with knees apart, hands on229 the floor; sitting on the floor with knees opened to one side; lying on her230 side in bed, propped on an elbow, looking back over the shoulder; kneeling with231 hands behind; reclining on a sofa with legs open.232- **L2 bold** (author style — always pair with the NSFW LoRA + low angle):233 standing with legs apart, shot from a low angle looking up; lying prone with234 hips lifted toward the camera; sitting with legs wide open, leaning forward on235 the elbows; kneeling on one knee with the other leg extended to the side, torso236 arched back; lying on her back with legs open, one hand behind the head.237- **L3 extreme / cropped**: extreme low angle looking up between open legs;238 overhead shot lying flat with legs open; cropped framing (waist-up or239 mid-thigh) to emphasize the pose.240241### 4.2 Suggestion table (user request → tier → concrete pose)242243| User asks for (examples) | Tier | Route ④ library pose families (`--list-poses` names) | Route ① prompt wording |244| --- | --- | --- | --- |245| 全着装、自然/编辑感 (standing, portrait, office/street) | L0 | Standing / Sitting / Walking / Close Up | standing with weight on one leg, hand in pocket; seated on a chair, legs crossed at the ankles |246| 泳装/湿身/海滩/浴室 (swimwear, wet, relaxed) | L1 | Squatting / Kneeling / Laying Back | crouching with knees apart, hands on the floor, looking back over the shoulder |247| 大胆/NSFW (nude + LoRA, low angle, "author style") | L2 | Kneeling / Laying Back / Prone / Splits / Crawling / Dancing on knees | standing with legs apart from a low angle; prone with hips lifted toward the camera; seated wide, leaning forward on elbows |248| 极端/特写构图 (crop, extreme low, overhead) | L3 | Splits / any L2 family + crop framing | extreme low angle looking up between open legs; overhead, lying flat with legs open |249250Rules:2511. Decide the reveal first: clothed → L0; swimwear/wet → L1; nude + NSFW LoRA → L2.252 **Pairing the LoRA with an L0 pose is the most common cause of "bland" NSFW253 output** (observed, not a proven law).2542. Camera angle amplifies the pose (empirical heuristic): low angle ≈ +1255 boldness, overhead ≈ +1. "Low angle + L1" ≈ "eye level + L2" — this is why256 the author's shower/massage shots read as bold despite simple poses.2573. One pose per prompt, limb by limb; two competing pose descriptions get blended.2584. When the user names a pose they saw somewhere, search the library first:259 `python3 scripts/run_vnccs_ui.py --list-poses` (136 entries with file paths).260261## 5. Route ② — Edit (pose change with identity preservation)262263```bash264python3 scripts/qwen_pose_test.py \265 --mode edit --name pose_change --input /absolute/base.png \266 --prompt "change her pose to kneeling on one knee, hands on hips, keep face hair and outfit" \267 --cfg 3 --sampler er_sde --scheduler beta --steps 20 \268 --out outputs/pose_change.png269```270271CFG 3 is more stable for edit mode. SFW instruction prompts keep NSFW LoRAs from272triggering; add `--lora qwen21-nsfw-image-edit.safetensors` only when undressing.273274## 6. Route ③ — Scripted pose transfer (3D mannequin reference)275276```bash277python3 scripts/run_pose_transfer.py \278 --char /absolute/character.png \279 --pose-ref /absolute/doll_render.png \280 --out outputs/pose_transfer.png [--seed N] [--steps 25]281```282283The pose source must be a textureless **3D mannequin render** — e.g. the284`doll_*.png` captures that route ④ saves under the ComfyUI output dir, or any2853D poser render of the target pose. Photos and illustrations as pose references286do NOT work: the model composites the reference person as a second person, or287redraws the reference person entirely (retested 2026-10-16 — all variants288failed).289290Verified recipe (2026-10-16) — the script sets it up automatically:291292- dual-image edit with **image 1 = mannequin** (pose) and **image 2 = character** (identity);293- prompt `replace the pose of <image 2> with the pose of <image 1>. keep the character of <image 2>`;294- **VNCCS PoseStudio LoRA @ 1.0 is mandatory** — without it the mannequin is295 ignored entirely (3 no-LoRA controls);296- 25 steps, cfg 1, euler/simple, 1024 px. Single person, identity kept, pose297 follows the render; stable across seeds (sit ×3 seeds + crawling /298 kneel-side / lean-back all passed). Camera framing can differ slightly from299 the render; producing *new* mannequin renders still requires a route ④300 browser session — route ③ reuses captures you already have.301302## 7. Route ④ — VNCCS PoseStudio (exact pose from the 3D library)303304The `VNCCS_PoseStudio` node needs the ComfyUI **web frontend** for its 3D "live305capture"; the headless API fails with "Live capture timed out". So the runner306drives a CDP browser: load the official graph → set character image + pose307JSON → queue → collect outputs from disk. ~60–90 s each.308309```bash310# list all 136 poses with resolved file paths311python3 scripts/run_vnccs_ui.py --list-poses312313# run one pose on a character image (+ optional scene prompt)314python3 scripts/run_vnccs_ui.py --pose "squatting leaning back" \315 [--char /absolute/face.png] [--prompt "on a rooftop at night, city bokeh"] [--timeout 900]316```317318- `--pose` takes the name/slug case-insensitively (`crawling`, `Squatting319 Leaning Back`, `arabesque`). 9 names exist in more than one repository320 (`Standing A`, `CloseUp A`, …); qualify with `<repo>/<category>/<name>`321 (exact keys printed by `--list-poses`, e.g.322 `miuproject_vnccs_poselibrary_main/standing/standing_a`).323- **Scene prompt (`--prompt`):** the capture prompt is rebuilt at execution324 time from the scene document's `export` block (default template325 `Draw character from image2\n<lighting>\n<user_prompt>`); `--prompt "..."`326 injects `export.user_prompt` before the graph loads. Editing the327 `TextEncodeQwenImage21` widget or adding `export.user_prompt` to a *library328 pose file* has no effect. To change the scene of an existing route ④ result,329 re-run it through route ② (`qwen_pose_test.py --mode edit`).330- The runner reaches a CDP browser through the OpenClaw adapter when present,331 otherwise through the bundled `scripts/browser_fallback.py` (reuses a332 reachable `COMFY_BROWSER_CDP` endpoint, else launches a throwaway headless333 Chromium; `COMFY_BROWSER_MODE=auto|attach|launch`).334- The runner re-indexes the pose library on every call — newly downloaded poses335 are usable immediately. Each run writes `<pose>_<stamp>.meta.json` next to336 the outputs and never clobbers previous results (unique comfy prefix per run).337- Output: `./outputs/vnccs_ui/` (override with `VNCCS_UI_OUT`). The 3D doll is a338 bare MakeHuman mannequin — nude prompts + pose transfer can produce nude339 output (expected: the model reads the doll's state).340- **Custom poses from code:** `scripts/make_custom_pose.py` loads a library341 pose into the PoseStudio editor, adds bone rotations, and saves through the342 editor's own Save (JSON + preview in343 `PoseLibrary/local_user_poses/poses/<category>/`). The new pose works with344 `run_vnccs_ui.py --pose` at once. `--list-bones` prints the base pose's345 euler angles; pass the full library id when a name is not unique.346347 ```bash348 python3 scripts/make_custom_pose.py \349 --base "MIUProject__VNCCS_PoseLibrary_Main/poses/Standing/Standing A" \350 --name "Arms Out" --rotate upperarm_l:1:90 --rotate upperarm_r:1:-90351 ```352- **Pose fidelity caveat:** each pose JSON carries its own `cameraParams`353 (yaw/pitch/zoom); the generated image inherits that viewpoint. Poses whose354 shape only reads from another angle (e.g. a true arabesque's extended back355 leg) come out front-facing and ambiguous — rotate the doll in the UI and356 re-run, or pick a pose that reads from the front.357358## 8. Environment & Safety359360- ComfyUI as a service (original host: `systemctl --user status|restart361 comfyui.service`). After installing new custom nodes, one restart is required.362- The route scripts take the bundled media lock in shared mode363 (`media_lock(exclusive=False, wait_timeout=...)`) so they don't fight video364 jobs on a shared machine; route ④ wraps the whole browser run in it.365- On the original host, video tasks hold the lock exclusively (~06:45 / 09:00 /366 11:30 openclaw windows); expect waits, not conflicts.367- If a graph includes `QwenImage21Cache` and the host errors on it368 (`aimdo memory compile error`), remove the node and rewire LoRA → KSampler369 (the route ④ runner does this automatically in the browser).370- Never modify the workflow JSONs in this package's `workflows/` while a run is371 in flight; the harness works on in-memory copies anyway.372- A 50-step 1152×1728 run takes ~200–400 s on the int8 model; a 30-step edit373 run ~60–120 s.374375## 9. Verification3763771. Check the harness printed `ok: true` and the output file is non-empty.3782. Visually inspect with an image-reading tool (composition, pose match,379 anatomy, face). For pose-transfer runs, compare against the reference.3803. For reproductions of published images, compare side-by-side with the author381 original (`downloads/samples/*_author.jpg`; comparison-sheet382 example: `scripts/build_phase5_sheet.py`).3834. Record prompt/seed/steps/LoRA in the run's `*.json` meta file (harnesses384 write one automatically for routes ①②③④).3855. When a recipe deviation was made (weaker prompt, no LoRA, fewer steps,386 default seed), label the output and state the deviation next to it in the387 delivery message.388389Last full verification: 2026-10-12 — route ① (4 author-prompt reproductions),390route ② (9/9 pose edits with identity preserved), route ④ (6 pose runs:391crawling, kneeling side, leaning back, sitting side, squatting leaning back,392arabesque — the last one verified for output only, pose-shape match unverified;393its doll capture turned out to be a stale legacy capture),394pose index count 136 == `GET /vnccs/pose_library/list`, pose-ladder A/B images395in `references/pose_examples/`.396397Follow-up checks 2026-10-14: route ④ `--prompt` (scene follows the prompt, pose398preserved), bundled browser fallback launch path, and custom pose creation via399`scripts/make_custom_pose.py` (save + output verified).400401Follow-up checks 2026-10-16 (route ③): 6 mannequin-reference transfers passed402(sit ×3 seeds, crawling, kneel-side, lean-back): single person, identity kept,403pose follows the render; 3 no-LoRA controls ignored the mannequin; photo and404illustration references still fail (composite / redraw of the reference405person). Script `scripts/run_pose_transfer.py` added and end-to-end tested.