Trellis.2
PulseAugur coverage of Trellis.2 — every cluster mentioning Trellis.2 across labs, papers, and developer communities, ranked by signal.
- 2026-06-17 product_launch A native MLX port of TRELLIS.2 was released for Apple Silicon. source
4 day(s) with sentiment data
-
Hugging Face releases guide for fine-tuning 3D generation model TRELLIS.2
Hugging Face has released a comprehensive guide detailing how to fine-tune TRELLIS.2, a 3D generation model. The guide focuses on using LoRA (low-rank adaptation) to improve specific aspects of 3D model generation, such…
-
JoVox framework enhances voxel generation with joint hierarchical flow matching
Researchers have developed JoVox, a novel discrete flow matching framework for generating high-fidelity yet lightweight voxel structures. Unlike previous hierarchical methods that make irreversible early decisions, JoVo…
-
AnchorComplete reconstructs complete 3D scene meshes from partial point clouds
AnchorComplete is a new framework designed to reconstruct complete 3D scene meshes from incomplete point cloud data. It addresses the common issue of partial reconstruction by integrating partial-point conditioning with…
-
StableGen 0.4.0 adds AI-powered 3D model generation and texturing
StableGen, an open-source Blender add-on, has released version 0.4.0, enabling users to generate and texture 3D models using local AI models via ComfyUI. This update introduces features like prompt-to-mesh generation wi…
-
Gate3DGen model purifies real-world images for improved 3D generation
Researchers have developed Gate3DGen, a novel pre-processing model designed to enhance the accuracy of 3D generation from real-world images. This model selectively identifies and purifies degraded images, such as those …
-
AI-generated images transformed into 3D prints with free tools
A new method allows users to transform AI-generated images into physical 3D prints using readily available free software. The process involves generating an image with Qwen Image 2.1, converting it into a 3D model using…
-
Microsoft's TRELLIS.2 image-to-3D model optimized for 16GB cards
A developer has successfully modified Microsoft's TRELLIS.2 image-to-3D model to run at its full 1536 resolution on a 16 GB graphics card, specifically an Nvidia GeForce RTX 5070 Ti. This was achieved by patching the mo…
-
New AI models advance 3D world reconstruction and modeling
Researchers have developed new methods for 3D world modeling and reconstruction from images. GeoState focuses on creating view-decoupled geometry states for predictive modeling, separating scene dynamics from camera var…
-
NOFACE3D framework enhances 3D asset detail from single images
Researchers have developed NOFACE3D, a novel framework designed to enhance the detail and resolution of 3D assets generated from single images. This training-free method focuses on refining critical regions, such as thi…
-
New method generates 3D scenes from single images, improving indoor and outdoor quality
Researchers have developed a new method for generating complete 3D scene meshes from a single image, extending existing object-centric generators like Trellis.2 to handle both indoor and outdoor environments. The approa…
-
New 3D Generation Methods Enable Part-Level Control and Selective Output
Researchers have developed Kaininja, a novel method that extends existing native 3D generators to produce part-level outputs. This is achieved through a dual-volume representation that addresses the challenge of represe…
-
Trellis.2 and Pixal3D 3D AI models now natively integrated into ComfyUI
Trellis.2 and Pixal3D, two significant models for 3D generative AI, are now natively integrated into ComfyUI. This integration eliminates the need for complex custom nodes, specific CUDA versions, or PyTorch downgrades,…
-
TRELLIS.2 INT8 ConvRot model runs natively on AMD RX 7900 XTX via ComfyUI
A developer has released a patch kit enabling the TRELLIS.2 INT8 ConvRot model to run natively on an AMD RX 7900 XTX GPU using ComfyUI. This implementation utilizes fused Triton kernels for improved performance, achievi…
-
User seeks open-source 2D-to-3D model workflow for 3D printing
A user on Reddit is seeking the best open-source, local workflow to convert 2D product images into editable, printable 3D models, specifically for creating hollow, watertight STLs with precise tolerances for lids. The u…
-
TRELLIS.2 simplifies 3D asset generation on low-VRAM GPUs
A new open-source tool called TRELLIS.2 has been developed to simplify the creation of high-quality 3D assets from images. This tool can generate a 3D asset in under 7 minutes, even on consumer-grade GPUs with as little…
-
User seeks help for accurate human 3D model generation with Trellis 2
A user on Reddit is seeking assistance with generating accurate 3D models of humans from images using the Trellis 2 software. While the software successfully creates detailed 3D models from various sources like AI-gener…
-
Open-source AI pipeline generates game assets locally
A developer has created a comprehensive pipeline for generating game assets using open-source AI models, ported to GGML for local execution. The pipeline includes tools for text-to-speech with voice cloning (OpenMOSS), …
-
Trellis 2 causes Mac crashes and setup issues for users
Users are encountering significant difficulties running Trellis 2 on Mac devices, with one user reporting extreme RAM usage exceeding 73GB and system crashes. Another user is seeking assistance with setting up Trellis o…
-
Trellis 2 workflow updated with Qwen 2.5 11B model
A user has updated the Trellis 2 multi-view workflow to integrate the Qwen 2.5 11B model. This update allows for better alignment of multi-view images before processing them with the Trellis 2 workflow. The updated work…
-
TRELLIS.2 ported to MLX for Apple Silicon, enabling image-to-3D generation
A native MLX port of Microsoft's TRELLIS.2 has been developed for Apple Silicon, enabling image-to-3D object model generation. This port focuses on usability for real-world workflows and supports resolutions of 512x512 …