fire/part-crafter 🔢🖼️❓✓ → 🖼️

▶️ 61 runs 📅 Aug 2025 ⚙️ Cog 0.16.1 🔗 GitHub ⚖️ License
game-asset-generation image-to-3d

About

PartCrafter is a structured 3D mesh generation model that creates multiple parts and objects from a single RGB image.

Example Output

Output

Example output

Performance Metrics

195.92s Prediction Time
195.93s Total Time
All Input Parameters 
{
  "seed": 0,
  "image": "https://replicate.delivery/pbxt/NSS5S3AQ8HNQi7dej7u0XMq06pv3gcbD8YAYNNClpZgRWpEd/out-1.webp",
  "num_parts": 16,
  "num_tokens": 2048,
  "guidance_scale": 7,
  "remove_background": false,
  "use_flash_decoder": false,
  "num_inference_steps": 50
}
Input Parameters
seed Type: integerDefault: 0Range: 0 - 10000
Random seed for reproducibility. Use 0 for random seed
image (required) Type: string
Input image for 3D mesh generation
num_parts Type: integerDefault: 16Range: 1 - 16
Number of parts to generate
num_tokens Default: 2048
Number of tokens for generation
guidance_scale Type: numberDefault: 7Range: 1 - 15
Guidance scale for generation
remove_background Type: booleanDefault: false
Remove background from input image
use_flash_decoder Type: booleanDefault: false
Use flash decoder for faster inference (Tempermental?)
num_inference_steps Type: integerDefault: 50Range: 10 - 100
Number of inference steps
Output Schema

Output

Type: string • Format: uri

Example Execution Logs 
Generating 16 parts with seed 3029
Denoising: 1/50
Denoising: 2/50
Denoising: 3/50
Denoising: 4/50
Denoising: 5/50
Denoising: 6/50
Denoising: 7/50
Denoising: 8/50
Denoising: 9/50
Denoising: 10/50
Denoising: 11/50
Denoising: 12/50
Denoising: 13/50
Denoising: 14/50
Denoising: 15/50
Denoising: 16/50
Denoising: 17/50
Denoising: 18/50
Denoising: 19/50
Denoising: 20/50
Denoising: 21/50
Denoising: 22/50
Denoising: 23/50
Denoising: 24/50
Denoising: 25/50
Denoising: 26/50
Denoising: 27/50
Denoising: 28/50
Denoising: 29/50
Denoising: 30/50
Denoising: 31/50
Denoising: 32/50
Denoising: 33/50
Denoising: 34/50
Denoising: 35/50
Denoising: 36/50
Denoising: 37/50
Denoising: 38/50
Denoising: 39/50
Denoising: 40/50
Denoising: 41/50
Denoising: 42/50
Denoising: 43/50
Denoising: 44/50
Denoising: 45/50
Denoising: 46/50
Denoising: 47/50
Denoising: 48/50
Denoising: 49/50
Denoising: 50/50
Decoding: 1/16
Decoding: 2/16
Decoding: 3/16
Decoding: 4/16
Decoding: 5/16
Decoding: 6/16
Decoding: 7/16
Decoding: 8/16
Decoding: 9/16
Decoding: 10/16
Decoding: 11/16
Decoding: 12/16
Decoding: 13/16
Decoding: 14/16
Decoding: 15/16
Decoding: 16/16
Inference time: 195.49 seconds
Generated 16 parts
Version Details
Version ID
a8fe4d91050b7d6497dd6676506a4105a3903b7287353681f52f3be0bb01d67f
Version Created
August 1, 2025
Run on Replicate →