fire/trellis – Run with an API on Replicate

fire / trellis

For the paper "Structured 3D Latents for Scalable and Versatile 3D Generation".

Cold

Public
210 runs
L40S
GitHub
Paper
License

Run with an API

Playground API Examples README Versions

Input

images

*file[]

Add multiple files

List of input images to generate 3D asset from

seed

integer

Random seed for generation

Default: 0

randomize_seed

boolean

Randomize seed

Default: true

generate_color

boolean

Generate color video render

Default: true

generate_normal

boolean

Generate normal video render

Default: false

generate_model

boolean

Generate 3D model file (GLB)

Default: false

save_gaussian_ply

boolean

Save Gaussian point cloud as PLY file

Default: false

return_no_background

boolean

Return the preprocessed images without background

Default: false

ss_guidance_strength

number

(minimum: 0, maximum: 10)

Stage 1: Sparse Structure Generation - Guidance Strength

Default: 7.5

ss_sampling_steps

integer

(minimum: 1, maximum: 50)

Stage 1: Sparse Structure Generation - Sampling Steps

Default: 12

slat_guidance_strength

number

(minimum: 0, maximum: 10)

Stage 2: Structured Latent Generation - Guidance Strength

Default: 3

slat_sampling_steps

integer

(minimum: 1, maximum: 50)

Stage 2: Structured Latent Generation - Sampling Steps

Default: 12

mesh_simplify

number

(minimum: 0, maximum: 0.98)

GLB Extraction - Mesh Simplification (only used if generate_model=True)

Default: 0.65

texture_size

integer

(minimum: 512, maximum: 2048)

GLB Extraction - Texture Size (only used if generate_model=True)

Default: 512

Run this model in Node.js with one line of code:

npx create-replicate --model=fire/trellis

or set up a project from scratch

Install Replicate’s Node.js client library:

npm install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import and set up the client:

import Replicate from "replicate";

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

Run fire/trellis using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "fire/trellis:ef5f7aacc4779906d75e891763af9db45f061ea382e263081e2b1ed38a5d72da",
  {
    input: {
      seed: 0,
      texture_size: 2048,
      mesh_simplify: 0,
      generate_color: false,
      generate_model: true,
      randomize_seed: true,
      generate_normal: false,
      save_gaussian_ply: false,
      ss_sampling_steps: 12,
      slat_sampling_steps: 12,
      return_no_background: false,
      ss_guidance_strength: 7.5,
      slat_guidance_strength: 3
    }
  }
);
console.log(output);

To learn more, take a look at the guide on getting started with Node.js.

Install Replicate’s Python client library:

pip install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import the client:

import replicate

Run fire/trellis using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "fire/trellis:ef5f7aacc4779906d75e891763af9db45f061ea382e263081e2b1ed38a5d72da",
    input={
        "seed": 0,
        "texture_size": 2048,
        "mesh_simplify": 0,
        "generate_color": False,
        "generate_model": True,
        "randomize_seed": True,
        "generate_normal": False,
        "save_gaussian_ply": False,
        "ss_sampling_steps": 12,
        "slat_sampling_steps": 12,
        "return_no_background": False,
        "ss_guidance_strength": 7.5,
        "slat_guidance_strength": 3
    }
)
print(output)

To learn more, take a look at the guide on getting started with Python.

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Run fire/trellis using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

curl -s -X POST \
  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Prefer: wait" \
  -d $'{
    "version": "ef5f7aacc4779906d75e891763af9db45f061ea382e263081e2b1ed38a5d72da",
    "input": {
      "seed": 0,
      "texture_size": 2048,
      "mesh_simplify": 0,
      "generate_color": false,
      "generate_model": true,
      "randomize_seed": true,
      "generate_normal": false,
      "save_gaussian_ply": false,
      "ss_sampling_steps": 12,
      "slat_sampling_steps": 12,
      "return_no_background": false,
      "ss_guidance_strength": 7.5,
      "slat_guidance_strength": 3
    }
  }' \
  https://api.replicate.com/v1/predictions

To learn more, take a look at Replicate’s HTTP API reference docs.

You can run this model locally using Cog. First, install Cog:

brew install cog

If you don’t have Homebrew, there are other installation options available.

Run this to download the model and run it in your local environment:

cog predict r8.im/fire/trellis@sha256:ef5f7aacc4779906d75e891763af9db45f061ea382e263081e2b1ed38a5d72da \
  -i 'seed=0' \
  -i 'texture_size=2048' \
  -i 'mesh_simplify=0' \
  -i 'generate_color=false' \
  -i 'generate_model=true' \
  -i 'randomize_seed=true' \
  -i 'generate_normal=false' \
  -i 'save_gaussian_ply=false' \
  -i 'ss_sampling_steps=12' \
  -i 'slat_sampling_steps=12' \
  -i 'return_no_background=false' \
  -i 'ss_guidance_strength=7.5' \
  -i 'slat_guidance_strength=3'

To learn more, take a look at the Cog documentation.

Run this to download the model and run it in your local environment:

docker run -d -p 5000:5000 --gpus=all r8.im/fire/trellis@sha256:ef5f7aacc4779906d75e891763af9db45f061ea382e263081e2b1ed38a5d72da
curl -s -X POST \
  -H "Content-Type: application/json" \
  -d $'{
    "input": {
      "seed": 0,
      "texture_size": 2048,
      "mesh_simplify": 0,
      "generate_color": false,
      "generate_model": true,
      "randomize_seed": true,
      "generate_normal": false,
      "save_gaussian_ply": false,
      "ss_sampling_steps": 12,
      "slat_sampling_steps": 12,
      "return_no_background": false,
      "ss_guidance_strength": 7.5,
      "slat_guidance_strength": 3
    }
  }' \
  http://localhost:5000/predictions

To learn more, take a look at the Cog documentation.

Output

model_file

no_background_image

{
  "completed_at": "2024-12-25T19:22:52.600278Z",
  "created_at": "2024-12-25T19:18:12.737000Z",
  "data_removed": false,
  "error": null,
  "id": "nbyyj1jr05rma0ckzx0t2kjea4",
  "input": {
    "seed": 0,
    "image": "https://replicate.delivery/pbxt/MCprdYbZYopM9oV25EeGmNQdKjeYjIB6sfyei5rTFeaXEoun/Set_in_role_playing_game__A_young_woman_with_realistic_adult_proportio_S1009161289_St8_G1.png",
    "texture_size": 2048,
    "mesh_simplify": 0,
    "generate_color": false,
    "generate_model": true,
    "randomize_seed": true,
    "generate_normal": false,
    "ss_sampling_steps": 12,
    "slat_sampling_steps": 12,
    "ss_guidance_strength": 7.5,
    "slat_guidance_strength": 3
  },
  "logs": "INFO:predict:Loading and preprocessing input image...\nINFO:predict:Saved image without background\nINFO:predict:Using randomized seed: 1505015317\nINFO:predict:Running TRELLIS pipeline...\nSampling:   0%|          | 0/12 [00:00<?, ?it/s]\nSampling:  17%|█▋        | 2/12 [00:00<00:00, 13.09it/s]\nSampling:  33%|███▎      | 4/12 [00:00<00:00, 13.14it/s]\nSampling:  50%|█████     | 6/12 [00:00<00:00, 13.16it/s]\nSampling:  67%|██████▋   | 8/12 [00:00<00:00, 13.21it/s]\nSampling:  83%|████████▎ | 10/12 [00:00<00:00, 13.18it/s]\nSampling: 100%|██████████| 12/12 [00:00<00:00, 14.37it/s]\nSampling:   0%|          | 0/12 [00:00<?, ?it/s]\nSampling:   8%|▊         | 1/12 [00:00<00:01,  7.12it/s]\nSampling:  25%|██▌       | 3/12 [00:00<00:00,  9.17it/s]\nSampling:  42%|████▏     | 5/12 [00:00<00:00,  9.66it/s]\nSampling:  58%|█████▊    | 7/12 [00:00<00:00,  9.89it/s]\nSampling:  75%|███████▌  | 9/12 [00:00<00:00,  9.99it/s]\nSampling:  92%|█████████▏| 11/12 [00:01<00:00, 11.01it/s]\nSampling: 100%|██████████| 12/12 [00:01<00:00, 10.70it/s]\nINFO:predict:TRELLIS pipeline complete!\nINFO:predict:Available output formats: dict_keys(['mesh', 'gaussian'])\nINFO:predict:Generating GLB model...\nINFO- Loaded 58336 vertices and 116560 faces.\n0% done\n2% done\n5% done\n7% done\n10% done\n13% done\n15% done\n18% done\n21% done\n23% done\n26% done\n28% done\n31% done\n34% done\n36% done\n39% done\n42% done\n44% done\n47% done\n50% done\n52% done\n55% done\n57% done\n60% done\n63% done\n65% done\n68% done\n71% done\n73% done\n76% done\n78% done\n81% done\n84% done\n86% done\n89% done\n92% done\n94% done\n97% done\n100% done\nRendering: 0it [00:00, ?it/s]\nRendering: 15it [00:00, 147.37it/s]\nRendering: 34it [00:00, 168.88it/s]\nRendering: 53it [00:00, 175.60it/s]\nRendering: 72it [00:00, 180.23it/s]\nRendering: 91it [00:00, 181.93it/s]\nRendering: 100it [00:00, 177.28it/s]\nINFO:predict:GLB model generation complete!\nINFO:predict:Prediction complete! Returning results...",
  "metrics": {
    "predict_time": 57.754129884,
    "total_time": 279.863278
  },
  "output": {
    "model_file": "https://replicate.delivery/xezq/68nSlbHf2Vw5NauxdzCCkCZf02071xM5xaEolEk6kfKZsB9nA/output.glb",
    "color_video": null,
    "normal_video": null,
    "combined_video": null,
    "no_background_image": "https://replicate.delivery/xezq/cUALxDQQdsIeCK4IOZNMfJ4KJxnemnTAO7diWJreVEAzYD6PB/output_no_background.png"
  },
  "started_at": "2024-12-25T19:21:54.846148Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/nbyyj1jr05rma0ckzx0t2kjea4",
    "cancel": "https://api.replicate.com/v1/predictions/nbyyj1jr05rma0ckzx0t2kjea4/cancel"
  },
  "version": "ace6e61c91a2674b520ef558d7e0d7494df803e82d2bdf51e8398788f51214b4"
}

Generated in

57.8 seconds

Tweak it ShareReport

INFO:predict:Loading and preprocessing input image...
INFO:predict:Saved image without background
INFO:predict:Using randomized seed: 1505015317
INFO:predict:Running TRELLIS pipeline...
Sampling:   0%|          | 0/12 [00:00<?, ?it/s]
Sampling:  17%|█▋        | 2/12 [00:00<00:00, 13.09it/s]
Sampling:  33%|███▎      | 4/12 [00:00<00:00, 13.14it/s]
Sampling:  50%|█████     | 6/12 [00:00<00:00, 13.16it/s]
Sampling:  67%|██████▋   | 8/12 [00:00<00:00, 13.21it/s]
Sampling:  83%|████████▎ | 10/12 [00:00<00:00, 13.18it/s]
Sampling: 100%|██████████| 12/12 [00:00<00:00, 14.37it/s]
Sampling:   0%|          | 0/12 [00:00<?, ?it/s]
Sampling:   8%|▊         | 1/12 [00:00<00:01,  7.12it/s]
Sampling:  25%|██▌       | 3/12 [00:00<00:00,  9.17it/s]
Sampling:  42%|████▏     | 5/12 [00:00<00:00,  9.66it/s]
Sampling:  58%|█████▊    | 7/12 [00:00<00:00,  9.89it/s]
Sampling:  75%|███████▌  | 9/12 [00:00<00:00,  9.99it/s]
Sampling:  92%|█████████▏| 11/12 [00:01<00:00, 11.01it/s]
Sampling: 100%|██████████| 12/12 [00:01<00:00, 10.70it/s]
INFO:predict:TRELLIS pipeline complete!
INFO:predict:Available output formats: dict_keys(['mesh', 'gaussian'])
INFO:predict:Generating GLB model...
INFO- Loaded 58336 vertices and 116560 faces.
0% done
2% done
5% done
7% done
10% done
13% done
15% done
18% done
21% done
23% done
26% done
28% done
31% done
34% done
36% done
39% done
42% done
44% done
47% done
50% done
52% done
55% done
57% done
60% done
63% done
65% done
68% done
71% done
73% done
76% done
78% done
81% done
84% done
86% done
89% done
92% done
94% done
97% done
100% done
Rendering: 0it [00:00, ?it/s]
Rendering: 15it [00:00, 147.37it/s]
Rendering: 34it [00:00, 168.88it/s]
Rendering: 53it [00:00, 175.60it/s]
Rendering: 72it [00:00, 180.23it/s]
Rendering: 91it [00:00, 181.93it/s]
Rendering: 100it [00:00, 177.28it/s]
INFO:predict:GLB model generation complete!
INFO:predict:Prediction complete! Returning results...

This example was created by a different version, fire/trellis:ace6e61c.

Examples

View more examples

Run time and cost

This model runs on Nvidia L40S GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

This model doesn't have a readme.