# Hunyuan 3D Generator

> Hunyuan 3D is Tencent's family of AI 3D generators. Kenerate AI's 3D Studio runs Hunyuan 3D 3.1 in its hosted Rapid edition, the studio's fastest engine: one front-view photo or a prompt of up to 200 characters becomes an untextured GLB in about a minute, made for drafts and blockouts. Version 3.1 was released in January 2026.

URL: https://kenerateai.com/model/hunyuan-3d
Publisher: Kenerate AI (https://kenerateai.com)
Last updated: 2026-10-01

## Key facts
- Engine here: Hunyuan 3D 3.1 Rapid (hosted)
- Released: 3.1: Jan 2026 · 3.0: Sep 2025 · 2.1 (open weights): Jun 2025
- Inputs: One front-view photo (≤8 MB) or a prompt up to 200 characters
- Output: Untextured (geometry-only) GLB
- Speed: About 1 minute — our test runs took 27–69 s
- Open source: Weights published up to 2.1; 3.x is hosted only
- Engine Meshy 7.1 (Meshy): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ; typical about 3–4 min; Production assets with the finest Meshy geometry (2K / 4K), smart topology, rigging and animation.
- Engine Meshy 7 (Meshy): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ; typical about 3–4 min; Detailed game and product assets, with an Ultra detail switch, low-poly mode and auto-rigging.
- Engine Meshy 6 (Meshy): inputs Text/Photo; files GLB, FBX, OBJ, USDZ; typical about 3–4 min; Reliable textured meshes with a polygon budget you control — the engine 3D Studio started on.
- Engine Tripo P2 (Tripo): inputs Text/Photo; files GLB, FBX; typical about 3–4 min; Crisp, game-ready meshes with pro texture tiers up to Extreme and clean quad remeshing.
- Engine Tripo H3.1 (Tripo): inputs Text/Photo/Multi-view; files GLB; typical about 4–5 min; High-quality meshes, with detailed geometry and a four-view mode for exact shapes.
- Engine GPT-6 Astra (OpenAI): inputs Text/Photo/Multi-view; files GLB; typical about 5–6 min; OpenAI's general GPT-6 Astra model run as a 3D job — follows long, detailed briefs and up to 10 reference photos, returns a GLB with PBR textures at real-world size. The slowest engine.
- Engine Rodin 2.5 (Hyper3D): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ, STL; typical about 2–3 min; Studio-grade meshes from 2K to 500K faces, quad or triangle, with a HighPack for 4K textures.
- Engine Rodin 2 (Hyper3D): inputs Text/Photo/Multi-view; files GLB, FBX, OBJ, USDZ, STL; typical about 2–3 min; Clean, detailed meshes — the first photo drives the materials.
- Engine Hunyuan 3D 3.1 Rapid (Tencent): inputs Text/Photo; files GLB; typical about 1 min; The fastest geometry — an untextured (clay) GLB for quick blockouts, drafts and ideas.

## Steps
1. Open 3D Studio — Pick Text to 3D or Image to 3D.
2. Pick Hunyuan 3D — Choose Hunyuan 3D 3.1 Rapid.
3. Add a photo or prompt — A front-on photo of one object, or up to 200 characters.
4. Generate — Inspect the clay model, then rebuild it textured on another engine if you like it.

## In depth

### Timeline
Hunyuan3D 2.1 was open-sourced in June 2025 [28]. Version 3.0 was unveiled in September 2025 together with Professional and Rapid editions of the cloud API; 3.1 followed in January 2026 with better fidelity and multi-view input on the Professional edition [30]. Tencent has since listed a 3.5 line.

### Open weights vs hosted
Only Hunyuan3D up to 2.1 (plus some side models) has public weights, under the Tencent Hunyuan 3D 2.1 Community License — which excludes the EU, UK and South Korea, needs a separate licence above 1 million monthly users and bans using outputs to train other models [29]. The 3.1 engine in the studio is a hosted service, so that licence governs the weights, not this engine.

### Why “Rapid” means untextured here
The Rapid endpoint we use takes one image or a short prompt and returns a mesh fast [35]. Our own test runs all came back as geometry with no materials or textures, which makes it ideal for drafts and printing tests.

References:
[1] Wikipedia — STL (file format): https://en.wikipedia.org/wiki/STL_(file_format)
[2] Wikipedia — Wavefront .obj file: https://en.wikipedia.org/wiki/Wavefront_.obj_file
[3] Wikipedia — FBX: https://en.wikipedia.org/wiki/FBX
[4] Khronos Group — glTF: https://www.khronos.org/gltf/
[5] 3MF Consortium — 3MF specification: https://3mf.io/specification/
[6] Wikipedia — Universal Scene Description (USD / USDZ): https://en.wikipedia.org/wiki/Universal_Scene_Description
[7] Wikipedia — PLY (file format): https://en.wikipedia.org/wiki/PLY_(file_format)
[8] Khronos Group — glTF 2.0 specification: https://registry.khronos.org/glTF/specs/2.0/glTF-2.0.html
[9] Apple — AR Quick Look: https://developer.apple.com/augmented-reality/quick-look/
[10] three.js — documentation: https://threejs.org/docs/
[11] Prusa Knowledge Base — Modeling with 3D printing in mind: https://help.prusa3d.com/article/modeling-with-3d-printing-in-mind_164135
[12] Prusa Blog — lithophanes and filament transparency: https://blog.prusa3d.com/lets-play-with-filament-transparency-how-to-make-amazing-lifelike-color-photos-with-just-a-simple-color-change_104338/
[13] Polymers (2024) 16(20):2862 — FFF parameters and translucency of PLA: https://doi.org/10.3390/polym16202862
[14] Wikipedia — Lithophane: https://en.wikipedia.org/wiki/Lithophane
[15] W3C — CSS Values and Units Level 3 (absolute lengths): https://www.w3.org/TR/css-values-3/
[16] Adobe Substance 3D — OpenGL vs DirectX normal maps: https://experienceleague.adobe.com/en/docs/substance-3d/bakers/common-questions/what-is-the-difference-between-the-opengl-and-directx-normal-format
[17] Blender Manual — Normal Map node: https://docs.blender.org/manual/en/latest/render/shader_nodes/displacement/normal_map.html
[18] Wikipedia — Normal mapping: https://en.wikipedia.org/wiki/Normal_mapping
[19] SIL Open Font License: https://openfontlicense.org/
[20] Meshy Blog — Meshy 7.1 launch (10 Sep 2026): https://www.meshy.ai/blog/meshy-7-1-launch
[21] Meshy Blog — Meshy 7 and image-to-3D geometry alignment: https://www.meshy.ai/blog/meshy-7-image-to-3d-geometry-alignment
[22] Meshy Help — Meshy 7 vs Meshy 6 vs Meshy 6 Lite: https://help.meshy.ai/en/articles/15697060-meshy-7-vs-meshy-6-vs-meshy-6-lite-which-ai-model-should-you-use
[23] Tripo Developer Docs — Tripo H3.1 (v3.1-20260211): https://developers.tripo3d.ai/en/models/v3-1
[24] VoxelMatters — Tripo AI launches P2.0 with native quad meshes (Sep 2026): https://www.voxelmatters.com/tripo-ai-launches-p2-0-model-to-generate-native-quad-meshes-for-production-pipelines/
[25] Hyper3D — Rodin Gen-2.5 launch press release (May 2026): https://pressadvantage.com/story/94732-hyper3d-launches-rodin-gen-2-5-bringing-sculpt-level-detail-and-production-control-to-ai-3d-generati
[26] arXiv — BANG: Dividing 3D Assets via Generative Exploded Dynamics (SIGGRAPH 2025): https://arxiv.org/abs/2507.21493
[27] befores & afters — Deemos launches Rodin Gen-2 (Oct 2025): https://beforesandafters.com/2025/10/02/deemos-launches-rodin-gen-2-a-groundbreaking-generative-ai-for-intuitive-3d-creation/
[28] GitHub — Tencent Hunyuan3D 2.1 (open weights): https://github.com/Tencent-Hunyuan/Hunyuan3D-2.1
[29] Hugging Face — Tencent Hunyuan 3D 2.1 Community License: https://huggingface.co/tencent/Hunyuan3D-2.1/blob/main/LICENSE
[30] Tencent Cloud — Hunyuan 3D product changelog: https://cloud.tencent.com/document/product/1804/120694
[31] CNBC — OpenAI rolls out GPT-6 Astra (3 Sep 2026): https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html
[32] AWS Bedrock — OpenAI GPT-6 Astra model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
[33] VoxelMatters — GPT-6 Astra can turn photos into 3D models: https://www.voxelmatters.com/gpt-6-astra-can-turn-photos-into-3d-models-and-playable-environments/
[34] WaveSpeedAI — GPT-6 Astra image-to-3D endpoint: https://wavespeed.ai/models/openai/gpt-6-astra/image-to-3d
[35] WaveSpeedAI — Hunyuan 3D 3.1 image-to-3D Rapid endpoint: https://wavespeed.ai/models/wavespeed-ai/hunyuan-3d-v3.1/image-to-3d-rapid

## FAQ
Q: What is Hunyuan 3D?
A: Hunyuan 3D (Hunyuan3D) is Tencent's family of AI 3D generation models, with releases from 1.0 in 2024 to 3.x in 2025–2026.

Q: Which version does Kenerate use?
A: Hunyuan 3D 3.1 in its hosted Rapid edition — fast, geometry-only output. Tencent also offers other editions and has listed a newer 3.5 line.

Q: Is Hunyuan 3D open source?
A: Partly. Tencent published open weights up to Hunyuan3D 2.1 (June 2025) under the Tencent Hunyuan 3D 2.1 Community License. Versions 2.5, 3.0 and 3.1 are hosted services, not open weights.

Q: What does the open-source licence restrict?
A: The 2.1 community licence doesn't apply in the EU, UK or South Korea, requires a separate licence above 1 million monthly users, and forbids using outputs to improve other AI models. It covers the open weights, not the hosted 3.1 engine.

Q: Why is my model grey?
A: The Rapid edition returns geometry only. Rebuild on a textured engine for colour.

Q: Can I find Hunyuan 3D on Hugging Face?
A: The open 2.x weights are on Hugging Face; to use 3.1 without setup, run it in the 3D Studio.

Q: Can I use a Hunyuan 3D model in Blender?
A: Yes. Download the GLB and import it with File → Import → glTF 2.0. The Rapid edition's mesh is untextured clay, which suits sculpting, blockouts and print tests; rebuild it on Tripo, Meshy or Rodin if you need colour. Apply scale after import, since glTF uses metres.

Q: Do I need a Tencent account?
A: No — Hunyuan 3D runs inside Kenerate's 3D Studio on any paid plan or credit pack.

## Related pages
- Tripo AI: https://kenerateai.com/model/tripo-ai
- Meshy AI: https://kenerateai.com/model/meshy-ai
- Rodin AI: https://kenerateai.com/model/rodin-ai
- GLB to STL: https://kenerateai.com/glb-to-stl

## Sources
- Wikipedia — STL (file format): https://en.wikipedia.org/wiki/STL_(file_format)
- Wikipedia — Wavefront .obj file: https://en.wikipedia.org/wiki/Wavefront_.obj_file
- Wikipedia — FBX: https://en.wikipedia.org/wiki/FBX
- Khronos Group — glTF: https://www.khronos.org/gltf/
- 3MF Consortium — 3MF specification: https://3mf.io/specification/
- Wikipedia — Universal Scene Description (USD / USDZ): https://en.wikipedia.org/wiki/Universal_Scene_Description
- Wikipedia — PLY (file format): https://en.wikipedia.org/wiki/PLY_(file_format)
- Khronos Group — glTF 2.0 specification: https://registry.khronos.org/glTF/specs/2.0/glTF-2.0.html
- Apple — AR Quick Look: https://developer.apple.com/augmented-reality/quick-look/
- three.js — documentation: https://threejs.org/docs/
- Prusa Knowledge Base — Modeling with 3D printing in mind: https://help.prusa3d.com/article/modeling-with-3d-printing-in-mind_164135
- Prusa Blog — lithophanes and filament transparency: https://blog.prusa3d.com/lets-play-with-filament-transparency-how-to-make-amazing-lifelike-color-photos-with-just-a-simple-color-change_104338/
- Polymers (2024) 16(20):2862 — FFF parameters and translucency of PLA: https://doi.org/10.3390/polym16202862
- Wikipedia — Lithophane: https://en.wikipedia.org/wiki/Lithophane
- W3C — CSS Values and Units Level 3 (absolute lengths): https://www.w3.org/TR/css-values-3/
- Adobe Substance 3D — OpenGL vs DirectX normal maps: https://experienceleague.adobe.com/en/docs/substance-3d/bakers/common-questions/what-is-the-difference-between-the-opengl-and-directx-normal-format
- Blender Manual — Normal Map node: https://docs.blender.org/manual/en/latest/render/shader_nodes/displacement/normal_map.html
- Wikipedia — Normal mapping: https://en.wikipedia.org/wiki/Normal_mapping
- SIL Open Font License: https://openfontlicense.org/
- Meshy Blog — Meshy 7.1 launch (10 Sep 2026): https://www.meshy.ai/blog/meshy-7-1-launch
- Meshy Blog — Meshy 7 and image-to-3D geometry alignment: https://www.meshy.ai/blog/meshy-7-image-to-3d-geometry-alignment
- Meshy Help — Meshy 7 vs Meshy 6 vs Meshy 6 Lite: https://help.meshy.ai/en/articles/15697060-meshy-7-vs-meshy-6-vs-meshy-6-lite-which-ai-model-should-you-use
- Tripo Developer Docs — Tripo H3.1 (v3.1-20260211): https://developers.tripo3d.ai/en/models/v3-1
- VoxelMatters — Tripo AI launches P2.0 with native quad meshes (Sep 2026): https://www.voxelmatters.com/tripo-ai-launches-p2-0-model-to-generate-native-quad-meshes-for-production-pipelines/
- Hyper3D — Rodin Gen-2.5 launch press release (May 2026): https://pressadvantage.com/story/94732-hyper3d-launches-rodin-gen-2-5-bringing-sculpt-level-detail-and-production-control-to-ai-3d-generati
- arXiv — BANG: Dividing 3D Assets via Generative Exploded Dynamics (SIGGRAPH 2025): https://arxiv.org/abs/2507.21493
- befores & afters — Deemos launches Rodin Gen-2 (Oct 2025): https://beforesandafters.com/2025/10/02/deemos-launches-rodin-gen-2-a-groundbreaking-generative-ai-for-intuitive-3d-creation/
- GitHub — Tencent Hunyuan3D 2.1 (open weights): https://github.com/Tencent-Hunyuan/Hunyuan3D-2.1
- Hugging Face — Tencent Hunyuan 3D 2.1 Community License: https://huggingface.co/tencent/Hunyuan3D-2.1/blob/main/LICENSE
- Tencent Cloud — Hunyuan 3D product changelog: https://cloud.tencent.com/document/product/1804/120694
- CNBC — OpenAI rolls out GPT-6 Astra (3 Sep 2026): https://www.cnbc.com/2026/09/03/open-ai-astra-gpt-6-cyber.html
- AWS Bedrock — OpenAI GPT-6 Astra model card: https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-openai-gpt-6-astra.html
- VoxelMatters — GPT-6 Astra can turn photos into 3D models: https://www.voxelmatters.com/gpt-6-astra-can-turn-photos-into-3d-models-and-playable-environments/
- WaveSpeedAI — GPT-6 Astra image-to-3D endpoint: https://wavespeed.ai/models/openai/gpt-6-astra/image-to-3d
- WaveSpeedAI — Hunyuan 3D 3.1 image-to-3D Rapid endpoint: https://wavespeed.ai/models/wavespeed-ai/hunyuan-3d-v3.1/image-to-3d-rapid