zylim0702/qwen-multi-view-advanced 🔢❓📝🖼️✓ → 🖼️

▶️ 443 runs 📅 Sep 2025 ⚙️ Cog 0.15.1

About

Example Output

Output

Performance Metrics

32.19s Prediction Time

32.20s Total Time

All Input Parameters

{
  "seed": -1,
  "art_style": "chibi",
  "body_type": "",
  "cfg_scale": 7,
  "eye_color": "",
  "body_style": "full_body",
  "hair_color": "",
  "hair_style": "",
  "view_count": "1_view",
  "accessories": "",
  "custom_pose": "",
  "input_image": "https://replicate.delivery/pbxt/NgPjy4TUz8Z1iEU3BykADWUIeJ7oKtvJnshBviuqtFodfjhH/portrait-young-asia-lady-with-positive-expression-arms-crossed-smile-broadly-dressed-casual-clothing-looking-camera-pink-background.jpg",
  "max_retries": 1,
  "target_size": 768,
  "detail_level": "high",
  "clothing_type": "",
  "clothing_color": "",
  "holding_objects": "",
  "art_style_custom": "",
  "clothing_pattern": "",
  "additional_prompt": "cartoon 3d rendering full body",
  "clothing_material": "",
  "facial_expression": "",
  "printing_friendly": true,
  "num_inference_steps": 10,
  "preserve_input_pose": true,
  "consistency_threshold": 0.7,
  "maintain_aspect_ratio": true,
  "negative_prompt_extra": "wrong body ratio"
}

Input Parameters

seed Type: integerDefault: 42: Random seed for reproducible results (set to -1 for random)
art_style Default: 3d_object: Art style preset
body_type Type: stringDefault:: Body type (e.g., 'muscular', 'slim', 'tall', 'short')
cfg_scale Type: numberDefault: 2.5Range: 1 - 10: CFG scale (1.0-10.0, higher = follows prompt more closely)
eye_color Type: stringDefault:: Eye color (e.g., 'blue', 'brown', 'green', 'red')
body_style Default: same_as_input: Body style (same as input image or full body view)
hair_color Type: stringDefault:: Hair color (e.g., 'blonde', 'black', 'red', 'blue')
hair_style Type: stringDefault:: Hair style (e.g., 'long', 'short', 'curly', 'ponytail')
view_count Default: 4_views: Number of views to generate (1 view = front only, 4 views = all angles)
accessories Type: stringDefault:: Accessories to add (comma-separated, e.g., 'hat, glasses', 'crown, earrings')
custom_pose Type: stringDefault:: Custom pose description (e.g., 'heroic stance', 'sitting', 'fighting pose')
input_image (required) Type: string: Input image to transform into 4 views
max_retries Type: integerDefault: 3Range: 1 - 5: Maximum number of retry attempts for low consistency
target_size Type: integerDefault: 1024Range: 512 - 2048: Target size for longer dimension when maintaining ratio (512-2048)
detail_level Default: high: Detail level
clothing_type Type: stringDefault:: Type of clothing (e.g., 'armor', 'dress', 'suit', 'casual clothes')
clothing_color Type: stringDefault:: Clothing colors (e.g., 'red and blue', 'gold and silver')
holding_objects Type: stringDefault:: Objects to hold (comma-separated, e.g., 'sword, shield', 'magic wand')
art_style_custom Type: stringDefault:: Custom art style description (only used if art_style='custom')
clothing_pattern Type: stringDefault:: Clothing pattern (e.g., 'stripes', 'geometric', 'plain')
additional_prompt Type: stringDefault:: Additional details to add
clothing_material Type: stringDefault:: Clothing material (e.g., 'metal', 'fabric', 'leather')
facial_expression Type: stringDefault:: Facial expression (e.g., 'smiling', 'serious', 'angry')
printing_friendly Type: booleanDefault: true: Optimize for 3D printing
num_inference_steps Type: integerDefault: 50Range: 10 - 200: Number of inference steps (10-200, higher = better quality but slower)
preserve_input_pose Type: booleanDefault: false: Keep the exact same pose as the input image (overrides all other pose options)
pose_reference_image Type: string: Reference image to extract pose from (overrides custom_pose)
consistency_threshold Type: numberDefault: 0.7Range: 0 - 1: Minimum consistency score required (0.0-1.0, lower = more retries)
maintain_aspect_ratio Type: booleanDefault: true: Preserve original image aspect ratio
negative_prompt_extra Type: stringDefault:: Things you specifically don't want

Output Schema

Output

Type: array • Items Type: string • Items Format: uri

Example Execution Logs

🎯 Starting Qwen 4-View Generation
⚡ Steps: 10, CFG: 7.0, Size: 768
🎨 Art Style: chibi
⚠️ Large image detected, will be resized for processing
✅ Input validation passed: (6000, 4000)
✅ Input image loaded successfully: (6000, 4000)
📐 Original image: 6000x4000 (aspect ratio: 1.50)
🔧 Configuring pipeline...
✅ Aspect ratio mode: True, target size: 768
✅ Generation config updated: steps=10, cfg=7.0
✅ Pipeline configuration completed
🔧 Processing customization inputs...
✅ Preserving input image pose
✅ Art style applied: chibi
✅ Additional prompt: cartoon 3d rendering full body
✅ Negative prompt: wrong body ratio
✅ 3D printing optimizations enabled
🔧 Applied 5 customizations:
- Preserve input pose
- Art style: chibi
- Additional details: cartoon 3d rendering full body
- Negative prompt: wrong body ratio
- 3D printing optimizations enabled
INFO:core.qwen_pipeline:✅ Aspect ratio preservation enabled (target size: 768px)
INFO:core.qwen_pipeline:Updated generation config: {'strength': 0.95, 'cfg_scale': 7.0, 'seed': 776688416363488, 'num_inference_steps': 10, 'true_cfg_scale': 4.0}
INFO:core.qwen_pipeline:🖼️  Resizing from 6000x4000 to 768x512 (aspect ratio: 1.50)
🎨 Generating front view only...
⏳ This should be faster with single view generation...
📊 Progress will be shown for each generation step
🔍 Extracting character description...
WARNING:core.qwen_pipeline:Vision processing method failed: No working vision processing method found, using manual extraction
ERROR:core.qwen_pipeline:Error extracting character description: list index out of range
INFO:core.qwen_pipeline:Base character description: 3D cartoon character with detailed features suitable for printing...
INFO:core.qwen_pipeline:🎨 Generating front view only...
INFO:core.qwen_pipeline:🔍 Received customization - pose: 'preserve the exact same pose and body positioning as shown in the input image', pose_source: 'preserve_input'
INFO:core.qwen_pipeline:👕 No clothing customization provided
INFO:core.qwen_pipeline:💇 No hair customization provided
INFO:core.qwen_pipeline:🏃 Using full body style
INFO:core.qwen_pipeline:🎭 Preserving input image pose: preserve the exact same pose and body positioning as shown in the input image...
INFO:core.qwen_pipeline:✅ Prompt polished: 'IMPORTANT: Transform this character to show the fr...' -> 'IMPORTANT: Transform this character to show the fr...'
INFO:core.qwen_pipeline:Pipeline does not support 'strength'; retrying without it
✅ Character description extracted successfully
🎨 Generating front view only...
✅ Front view generated successfully
⚠️ Missing views: ['left', 'back', 'right']
✅ Generated 1 views successfully
✅ Single view generated successfully (no consistency check needed)
💾 Saving generated views...
INFO:core.qwen_pipeline:✅ Front view generated successfully
✅ Saved front view: front_view.png
✅ All views saved successfully (1 files)
🎉 Generation completed!
📊 Consistency Score: 1.00/1.0
📁 Generated 1 files

Version Details

Version ID: c104b08d498d9a8eedbe0d1feb68f87448b640e0521be8b0264a479193f4167a
Version Created: September 14, 2025

Run on Replicate →