3D
IDOL Image to 3D Splat
IDOL: one picture of an object to a 3D Gaussian splat PLY.
Built with DINOv3.
Copyright (c) 2026 VAST
模型授權
IDOL · Apache-2.0
Apache License 2.0. Keep the licence text with the weights. Card declares apache-2.0; the repo publishes no LICENSE file, so the standard text is shipped.
DINOv3 ViT-H+/16 · LicenseRef-DINOv3-License
Redistribution under the DINOv3 License with a copy of it; use must comply with trade-control law (no ITAR, military, nuclear or weapons end uses); no reverse engineering. Research publications must acknowledge DINO Materials. The text sets no commercial-use limit and no Built-with label; the label is Mesh's choice.
BiRefNet · MIT
MIT, Copyright (c) 2024 ZhengPeng. Keep the copyright and permission notice. The HF repo has no LICENSE file; the text is from the GitHub repo.
FLUX.2-klein-4B VAE encoder · Apache-2.0
Apache License 2.0. Keep the licence text with the weights. Only the encoder is kept.
TripoSplat (probable IDOL denoiser ancestor) · MIT
Conservative notice preservation for probable lineage, not a claim of checkpoint identity. Keep VAST copyright and MIT permission notice.
API 參考
透過 REST 提交
使用您的個人 API 金鑰(msk_…),從您自己的工具為此工作負載建立工作。以這種方式啟動的工作,會在您於 您的裝置 連接的同一批裝置上執行——至少須有一部在線上。
https://cp.meshlabs.cloud/v1/organizations/jobs使用 X-Organization-Api-Key: msk_your_personal_api_key 進行驗證。此標頭名稱與組織金鑰相同;個人金鑰使用同一個端點。
POST https://cp.meshlabs.cloud/v1/organizations/job-inputs/presign,將您的檔案上傳到傳回的 upload_url,然後在 job_spec 中傳入 download_url(例如作為 input_image_url)。請求內文
{
"workload_type": "idol-i2gs",
"job_spec": {
"input_image_url": "https://…/your-input.png",
"mode": "synthetic",
"steps": 50,
"cfg_strength": 3,
"rescale_t": 3,
"num_gaussians": 1048576,
"seed": 0
}
}job_spec 欄位
| 欄位 | 必填 | 預設值 | 說明 |
|---|---|---|---|
input_image_url | 是 | — | The picture: an object on a transparent background (a PNG with its alpha). |
mode | 是 | — | Whether the picture is a synthetic render or a real photograph. It sets the crop margin. |
crop_margin | 否 | — | The crop around the object. Defaults to 1.2 for synthetic and 1.7 for real. |
steps | 否 | 50 | Denoising steps. |
cfg_strength | 否 | 3 | Guidance against an empty condition. 0 turns it off and halves the work. |
rescale_t | 否 | 3 | How far the time grid is shifted towards noise. |
num_gaussians | 否 | 1048576 | Gaussians in the splat, a multiple of 32. |
seed | 否 | 0 | Random seed. No seed is seed 0. |
cURL
curl -X POST https://cp.meshlabs.cloud/v1/organizations/jobs \
-H "X-Organization-Api-Key: msk_your_personal_api_key" \
-H "Content-Type: application/json" \
-d '{
"workload_type": "idol-i2gs",
"job_spec": {
"input_image_url": "https://…/your-input.png",
"mode": "synthetic",
"steps": 50,
"cfg_strength": 3,
"rescale_t": 3,
"num_gaussians": 1048576,
"seed": 0
}
}'https://cp.meshlabs.cloud/v1/organizations/jobs/{job_id}使用同一個 API 金鑰輪詢,直到 status 進入終止狀態(COMPLETED、FAILED、CANCELED 或 MODERATION_BLOCKED)。啟用內容審核時,工作可能會先處於 MODERATION_PENDING 狀態,之後才進入運算佇列。已完成的工作會在每個輸出上附帶預先簽署的 download_url 值。
建立回應
{
"id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
"workload_type": "idol-i2gs",
"status": "PENDING",
"job_spec_json": {
"input_image_url": "https://…/your-input.png",
"mode": "synthetic",
"steps": 50,
"cfg_strength": 3,
"rescale_t": 3,
"num_gaussians": 1048576,
"seed": 0
},
"work_units": [],
"outputs": [],
"created_at": "2026-06-08T12:00:00Z",
"updated_at": "2026-06-08T12:00:00Z"
}完成回應
{
"id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
"workload_type": "idol-i2gs",
"status": "COMPLETED",
"job_spec_json": {
"input_image_url": "https://…/your-input.png",
"mode": "synthetic",
"steps": 50,
"cfg_strength": 3,
"rescale_t": 3,
"num_gaussians": 1048576,
"seed": 0
},
"work_units": [
{
"id": "…",
"index": 0,
"status": "COMPLETED"
}
],
"outputs": [
{
"id": "…",
"output_type": "<output type>",
"filename": "<output filename>",
"size_bytes": 524288,
"download_url": "<presigned output download URL>"
}
],
"created_at": "2026-06-08T12:00:00Z",
"updated_at": "<completion timestamp>"
}Upload one picture of an object on a transparent background and download a 3D Gaussian splat as a PLY file. IDOL runs 50 denoising steps and up to 1,048,576 Gaussians.