Morpheus Fashion Design

Generate professional fashion/product advertising images using ComfyDeploy's Morpheus Fashion Design workflow.

⚠️ CRITICAL RULES

1. NO `logo` field — EVER

The logo input has been removed from the API. Do NOT pass it. Only two image inputs exist:

- product → the product being advertised
INLINECODE3 → frontal face photo of the model

2. NEVER USE AUTO VALUES for packs

Configuration packs MUST NEVER be left on auto or AUTO.

INLINECODE6 = empty values = neutral, boring images with no creative direction.

Always select deliberately based on the brief. Custom string values are allowed and encouraged.

Pack Selection Guidelines

Pack	How to Choose
INLINECODE7	Brand personality: luxury→`premium_restraint`, sports→`cinematic_realism`, street→INLINECODE10
INLINECODE11

Sports→sony_a1, editorial→hasselblad_x2d, street→leica_m6 | | lens_pack | Portrait compression? Wide? Match shot type and mood | | lighting_pack | Golden hour? Studio? Natural window? Match brief | | pose_discipline_pack | Sport action→sport_in_motion, commercial→commercial_front_facing, UGC→street_style_candid_walk | | film_texture_pack | Warm editorial→kodak_portra_400, cinematic→kodak_vision3_500t, clean→digital_clean_no_emulation | | environment_pack | beach_minimal, urban_glass_steel, street_crosswalk — or custom string | | color_science_pack | warm_golden_editorial, neutral_premium_clean, cinematic_low_contrast | | time_weather_pack | golden_hour_clear, bright_midday_sun, overcast_winter_daylight |

API Details

Endpoint: https://api.comfydeploy.com/api/run/deployment/queue
Deployment ID: INLINECODE38

Canonical API Call

CODEBLOCK0

🎭 Model Catalog

GitHub: https://github.com/PauldeLavallaz/model_management
Local path: INLINECODE40

Priority order

1. User provides model image → use directly
User describes model → search catalog, select best match
No specification → choose based on brief context

Searching

CODEBLOCK1

Creative Brief — How to Write It

Write it as a real photography director's brief — specific, physical, cinematic:

Example: CasanCrem Light (ironic/viral UGC angle)

CODEBLOCK2

Example: Oakley Snowboarding

CODEBLOCK3

Configuration Packs Reference

Pack	Options
INLINECODE41	INLINECODE42, `editorial_precision`, `cinematic_realism`, `cinematic_memory`, `campaign_hero`, `product_truth`, `clean_commercial`, `street_authentic`, `archive_fashion`, INLINECODE51
INLINECODE52

full_body_wide, medium_shot, close_up, low_angle_hero, three_quarter, waist_up | | camera_pack | arri_alexa35, canon_r5, hasselblad_x2d, leica_m6, sony_a1 | | lens_pack | cooke_anamorphic_i_50, leica_noctilux_50, zeiss_otus_55, wide_distortion_controlled | | lighting_pack | golden_hour_backlit, natural_window, studio_three_point, bright_midday_sun | | pose_discipline_pack | commercial_front_facing, street_style_candid_walk, sport_in_motion | | film_texture_pack | kodak_portra_400, fujifilm_velvia_50, kodak_vision3_500t, digital_clean_no_emulation | | color_science_pack | neutral_premium_clean, warm_golden_editorial, cinematic_low_contrast | | environment_pack | beach_minimal, urban_glass_steel, street_crosswalk — or custom descriptive string | | time_weather_pack | golden_hour_clear, bright_midday_sun, overcast_winter_daylight | | branding_pack | logo_none ← always unless logo explicitly requested | | intent | awareness, consideration, conversion, retention | | aspect_ratio | 9:16, 16:9, 1:1, 4:5, 3:4 |

Python Upload Helper

CODEBLOCK4

Priority Hierarchy

CODEBLOCK5

Troubleshooting

Imagen negra/vacía = filtro de moderación de Google/Gemini activado.

- No usar personas famosas o celebridades
Modificar el brief eliminando referencias problemáticas

API Key: NO pasar como parámetro. Ya está configurado en ComfyDeploy.

Integración con Portrait Generator

Flujo de selección de modelo (actualizado)

El campo model puede venir de dos fuentes:

Opción A — Catálogo (default):

~/clawd/models-catalog/catalog/images/model_XX.jpg

Usar cuando el guión no requiere rasgos muy específicos.

Opción B — Portrait Generator (cuando el guión lo requiere):

portrait-generator → imagen_1 (vista frontal) → model en Morpheus

Usar cuando necesitás: etnia muy específica, edad exacta, rasgos únicos, personaje
que no existe en el catálogo de 114 modelos.

Regla: imagen_1 del Portrait Generator = siempre la vista frontal = la correcta para model.

Cuándo usar Portrait Generator

- Personaje mayor de 60 años
Etnia muy específica no disponible en catálogo
Rasgos físicos únicos (cicatrices, vitiligo, asimetría facial)
Cualquier personaje que el guión describa con detalle

Cuándo usar el catálogo

- Personaje genérico / sin especificaciones
Cualquier modelo joven-adulto estándar
Cuando la velocidad importa más que la precisión

Morpheus Fashion Design

使用ComfyDeploy的Morpheus Fashion Design工作流生成专业的时尚/产品广告图像。

⚠️ 关键规则

1. 绝对不要使用logo字段

logo输入已从API中移除。请勿传入。仅存在两个图像输入：

- product → 被广告的产品
model → 模特的正面面部照片

2. 切勿对配置包使用自动值

配置包绝不能保留为auto或AUTO。

auto = 空值 = 中性、无创意方向的乏味图像。

始终根据需求说明有目的地选择。允许并鼓励使用自定义字符串值。

配置包选择指南

配置包	如何选择
stylepack	品牌个性：奢侈品→premiumrestraint，运动→cinematicrealism，街头→streetauthentic
camerapack

API详情

端点： https://api.comfydeploy.com/api/run/deployment/queue
部署ID： 1e16994d-da67-4f30-9ade-250f964b2abc

标准API调用

javascript
const response = await fetch(https://api.comfydeploy.com/api/run/deployment/queue, {
method: POST,
headers: {
Content-Type: application/json,
Authorization: Bearer YOURAPIKEY
},
body: JSON.stringify({
deployment_id: 1e16994d-da67-4f30-9ade-250f964b2abc,
inputs: {
product: / 产品图片URL /,
model: / 模特面部URL /,
brief: 详细的场景、姿势、灯光、氛围、产品摆放...,
target: 目标受众：人口统计、心理特征、生活方式...,
input_seed: -1,
brandingpack: logonone,
aspect_ratio: 9:16,
stylepack: streetauthentic,
camerapack: sonya1,
lenspack: zeissotus_55,
filmtexturepack: kodakportra400,
colorsciencepack: warmgoldeneditorial,
shotpack: mediumshot,
posedisciplinepack: streetstylecandid_walk,
lightingpack: naturalwindow,
timeweatherpack: goldenhourclear,
environmentpack: urbanglass_steel,
intent: awareness
}
})
});

🎭 模特目录

GitHub： https://github.com/PauldeLavallaz/model_management
本地路径： ~/clawd/models-catalog/catalog/images/

优先级顺序

1. 用户提供模特图像 → 直接使用
用户描述模特 → 搜索目录，选择最佳匹配
无规格说明 → 根据需求说明上下文选择

搜索

bash

按体型

cat models-catalog/catalog/catalog.json | python3 -c import json,sys; data=json.load(sys.stdin) for t in data[talents]: if t.get(bodytype) in [curvy,plussize]: print(t[id], t[name], t[ethnicity])

按族裔+性别

cat models-catalog/catalog/catalog.json | jq [.talents[] | select(.ethnicity == hispanic and .gender == female) | {id, name, body_type}]

创意需求说明 — 如何编写

将其写成真正的摄影导演需求说明——具体、物理、电影感：

示例：CasanCrem Light（讽刺/病毒式UGC角度）

CasanCrem Light La Serenísima的UGC TikTok活动。
体格健壮的年轻人在阿根廷家庭厨房中，手捧产品罐，面带会意的微笑。
放松的姿势，靠在台面上，直视镜头，充满能量和默契。
产品罐在手中可见且突出。午后温暖的自然窗光。
真实的UGC风格，非传统广告。真实身材， relatable，非运动型。

示例：Oakley单板滑雪

Oakley单板滑雪活动。骑手在雪地公园的金属滑轨上，身体略微旋转，
手臂张开保持平衡，视线看向路线。技术性真实的自由式姿势。
高山雪地公园，正午充足日光，压实雪地，金属结构。
自然阳光在雪地上反射——饱和色彩，强对比。
纪实手法——定格骑手在滑轨上的瞬间，身体和滑板清晰锐利。
真实滑行场景：平衡、专注、风格融为一体。

配置包参考

配置包	选项
stylepack	premiumrestraint，editorialprecision，cinematicrealism，cinematicmemory，campaignhero，producttruth，cleancommercial，streetauthentic，archivefashion，experimentalauthorial
shotpack

Python上传辅助函数

python
def comfyupload(filepath: str, apikey: str) -> str:
from pathlib import Path
import requests
p = Path(filepath)
mime = image/png if p.suffix ==

morpheus-fashion-design墨菲斯时装设计