Skip to content

Fields

3D generation latest news

3D assets and scenes from text or images: meshes, materials, topology quality and whether they hold up in an engine.

79 picks52 in the last 30 days228 collected in all

Updated

Picks archive · Page 2

Nov 11Wed21–40
  1. 80 Level47

    Blender 5.3 ships with native 3D Gaussian Splatting and DLSS denoising

    Blender 5.3 is out, with native 3D Gaussian Splatting support: splats can be imported from PLY, SPZ and glTF files, processed with Geometry Nodes, and rendered directly in EEVEE or Cycles.

    Why it matters: The article details Blender 5.3's native support for 3D Gaussian Splatting and DLSS denoising, giving art pipelines what they need to judge whether to upgrade.

  2. fal44

    fal updates Meshy Rigging to turn humanoid 3D models into animatable characters

    fal has launched its Meshy Rigging update, which turns humanoid 3D models into characters that can be used directly for animation. The feature works with tools including Unreal Engine, Unity, Blender and Mixamo; finger rigging costs nothing extra, and multiple animations can be stacked on the same character.

    Why it matters: The update spells out which animations the rigged output supports and which engines it works with, so readers can judge whether their existing 3D assets can plug straight into an animation pipeline.

Nov 10Tue
  1. QbitAI57

    Fable 5.2's first hands-on tests: users build playable 3D games, 3D worlds and videos

    After Fable 5.2's release, early testers and researchers used it directly to generate playable 3D games, 3D worlds, videos and music; Wharton professor Ethan Mollick ran about 25,000 agents to generate more than 450 3D worlds.

    Why it matters: The piece rounds up hands-on cases, benchmarks and pricing data from several developers, so you can compare the new model's actual output and cost in game and 3D content generation.

Nov 9Mon
  1. Hyper3D Rodin46

    Hyper3D launches HYPER3D Rigging to turn a single design sheet into a riggable 3D character for Unreal Engine

    Hyper3D has launched HYPER3D Rigging, which automatically rigs skeletons for humans, animals and fantasy creatures down to individual fingers, and supports adding bones directly with prompts. The feature ships with 250+ motion presets and animation controls, and the official demo shows a complete riggable character generated from a single design sheet and then brought into Unreal Engine for rendering and control. Agentic Mode in HYPER3D Rodin is what takes a static character into a movable state.

    Why it matters: The demo shows the full path from one design sheet to a rigged character inside Unreal Engine, plus 250+ motion presets that can be applied directly.

  2. Hyper3D Rodin55

    Hyper3D opens its Rigging API to all users

    Hyper3D by Deemos announced that its Rigging API is now available to all users, with documentation at https://docs.hyper3d.ai/en/api-specification/rigging. The company previously described the feature as automatically rigging humans, animals and fantasy creatures, down to individual fingers, with bones added or removed via prompt. It also offers 250+ motion presets and animation controls, so rigging, adjustment and animation all happen on one platform.

    Why it matters: The Rigging API is now open to all users, and the accompanying documentation lists request parameters and skeleton template options, so readers can gauge the cost of wiring it into an existing 3D pipeline.

Nov 8Sun
  1. Grok Bot Research48

    Three-Live v1.3.1 released with MCP service for validating, rendering and simulating 3D scenes

    Three-Live v1.3.1 Beta adds a built-in MCP Stdio server that exposes scene validation, offscreen rendering and simulation to assistants such as Claude Code, Cursor, Gemini CLI and Windsurf.

    Why it matters: A built-in MCP service puts scene validation, offscreen rendering and physics simulation into the AI assistant's context, making it easy to judge whether it fits your own 3D scene workflow.

Nov 7Sat
  1. QbitAI49

    Moxin Technology releases general 4D world model MoWorld 2.0, with 70 FPS inference on Ascend NPUs

    On November 4, Moxin Technology released its next-generation general-purpose 4D world model MoWorld 2.0 at the World Internet Conference in Wuzhen, folding more than ten physical elements — including gravity, elasticity, buoyancy, lighting, soft bodies, fluids, rigid bodies and thermodynamics — into a single model.

    Why it matters: The original article lays out a unified modeling approach for more than ten physical elements, plus 70 FPS inference on Ascend NPUs and a 30% cost comparison against GPU setups, which readers can use to gauge the conditions for putting physics simulation into practice.

Nov 6Fri
  1. Hyper3D Rodin39

    HYPER3D Rigging launches, turning a static model into a fully animated character in 15 minutes

    Hyper3D by Deemos has released HYPER3D Rigging, which automatically rigs a static model into a fully animated character, with an official workflow time of 15 minutes. The tool auto-rigs humans, animals and fantasy creatures down to each finger, offers 250+ motion presets and animation controls, and supports prompt-based editing in the rigging step — for example, typing "Add tail bones." adds tail bones. The post says game developer @HOYO_MAX used it to finish character animation.

    Why it matters: The post lists the creature types covered by automatic rigging and 250+ motion presets, which you can weigh against your own project's character count and rigging workflow.

  2. 80 Level64

    Entire Quixel Megascans library is now permanently free on Fab, with terms opened to AI training

    Quixel has made the full Megascans library of roughly 23,000 3D scanned assets permanently free on Fab, with asset page updates continuing through November 12. The terms for all Quixel assets were updated at the same time to allow AI models to be trained on Megascans and Megaplants, so models can learn from real scans and generate photoreal environments.

    Why it matters: Unlike the limited-time 2024 giveaway, this one arrives together with terms that open AI training, so it can be used to judge whether scanned assets can enter a generative content pipeline.

  3. Grok Bot Research44

    Frontier3DBench benchmarks coding agents on 3D objects, scenes, and interactive open-world building with 145 tasks

    Xinjie Lin and colleagues propose Frontier3DBench, which uses 145 tasks across 9 categories to cover object, scene, and interactive open-world creation and evaluates appearance, structure, and behavior together. Among 11 model and harness configurations, GPT-6.1-sol leads at the object and scene levels with an overall score of 66.02/100, while GPT-6-astra leads on interactive world building; every configuration scores lower on reference reconstruction than on text-driven generation.

    Why it matters: Stratified scores across 11 configurations identify reference reconstruction and coordinated-activity completion as the main bottlenecks for coding agents working in 3D, giving a reference point for model selection and method comparison.

  4. arXiv Game AI71

    PlayGen2: Learning real-time playable worlds with explicit 3D state

    The authors propose PlayGen2, a real-time interactive world model that separates learned 3D state evolution from neural rendering, using a volumetric occupancy grid and an identity-indexed 3D state to maintain persistent world memory. Tested on Doom, the model learned world dynamics including geometry-aware player movement and player-NPC interaction, and the full simulation and rendering pipeline runs at about 160ms backend latency on a MacBook Pro with an M4 Pro chip.

    Why it matters: The paper splits a playable world into explicit 3D state evolution and neural rendering, and reports measured latency and zero-shot transfer results, making it easier to judge whether such world models can support real-time interaction.

  5. Grok Bot Research58

    HairDeck open-sources a model that generates stylized hair with UVs, colors and spring bones from a single image

    Researchers from Tsinghua University, CMU and The Chinese University of Hong Kong open-sourced HairDeck, which generates stylized hair cards with per-card UVs and colors from a single character image, along with the spring bone chains that drive them. The model is split into weights including Image DiT, Card VAE, Color DiT and Chain DiT, outputs GLB hair meshes, and can be attached to the provided VRoid Studio template character to export VRM 0.x avatars.

    Why it matters: The original post releases the weights of each module and a local running workflow, which readers can use to judge whether single-image hair asset generation can fit into their own character pipelines.

  6. fal47

    Meshy Relief and N1 Retexture are now live on fal

    fal has added Meshy Relief and N1 Retexture to its platform. N1 Retexture takes a single front view, or up to four views of the same object, and rebuilds a clean 2K base color texture for an existing 3D model. Meshy Relief turns a single photo into a 3D relief, ranging from the shallow relief suited to badges to sculptural high relief, and the exported relief has a flat base so it can be printed and displayed directly.

    Why it matters: Both tools need only a single image or a handful of views as input, so their integration cost can be assessed against existing 3D assets and 3D printing workflows.

Nov 5Thu
  1. Hyper3D Rodin52

    Hyper3D launches HYPER3D Rigging, an auto-rigging tool with prompt-based skeleton editing

    Hyper3D has launched HYPER3D Rigging, which automatically rigs skeletons for humans, animals and fantasy creatures down to individual fingers, with skeletons editable through prompts. The product description quoted lists 250+ motion presets and animation controls, and gives the example that entering "Add tail bones" adds tail bones, with rigging, adjustment and animation all done in one place.

    Why it matters: The source lays out the full chain from a static 3D model to an animated character, and its prompt-based skeleton editing can serve as a reference for animation previs.

  2. Youxituoluo55

    Hands-on: Kimi K3.1 Pro builds a game from a single prompt, completing modeling and importing it into a Godot level

    Youxituoluo ran a hands-on test of the Kimi K3.1 Pro desktop app, giving it a single instruction to complete scene and character modeling and then import the assets into the Godot engine to assemble a level with interaction points and mechanism animations. During character modeling, Kimi called Tripo3D directly, but the results across multiple generations deviated noticeably from the reference images; the author concludes that repeated back-and-forth correction is still required, and the artists interviewed also said AI-generated characters are not yet usable as-is.

    Why it matters: The author ran the entire pipeline from modeling to engine import on the lowest subscription tier and logged a spend of about 13.08 yuan, which serves as a reference point for evaluating this kind of workflow.

Nov 4Wed
  1. arXiv Game World Models & Generation43

    Paper proposes DRE method using wireframe and UV seam views to judge 3D asset usability

    Researchers propose Diagnostic Rendered Evidence (DRE), which feeds three aligned views — beauty, wireframe, and UV seams plus wireframe — to separate ViT classifiers frozen for each purpose and then fuses their scores in a late fusion step, in order to judge whether a 3D asset can be reliably reused in scene assembly, editing, rendering, and game production.

    Why it matters: The paper uses wireframe and UV seam views to cover the blind spots of judging from the beauty render alone, and reports 86.36% balanced accuracy on 1,234 manually annotated assets.

  2. Grok Bot Research53

    Kimi releases game development and creative work solution with Unity, Unreal Engine and Blender plugins

    Kimi released its game development and creative work solution, built on the Kimi K3.1 series models and the Kimi Code and Kimi Work series agent products and connecting to professional software including Unity, Unreal Engine and Blender through plugins to operate existing projects.

    Why it matters: The article lays out the split of work across prototype validation, asset adaptation and in-engine changes, which you can compare against your own pipeline to judge which stages suit being handed to an agent.

Nov 3Tue
  1. Meshy44

    Meshy releases N1 Texture texturing model, texturing a single game asset in about 20 seconds

    Meshy released the N1 Texture texturing model, saying a single game asset takes about 20 seconds to texture, with clean 2K base color output and fewer seams, aimed at stylized game assets. In the demo, the team used GPT to write a store, Meshy T2 to convert product photos into 3D, and N1 to texture products from apples to donuts, opening a supermarket in 10 minutes. The tweet ends by asking readers what they would open in 10 minutes.

    Why it matters: The post gives a per-asset texturing time of about 20 seconds and 2K base color output, which you can use to judge whether it fits into a stylized project's asset pipeline.

  2. arXiv Game World Models & Generation34

    Isekai3D SDG proposes a synthetic gameplay visual data platform built on an Unreal Engine plugin

    A research team proposes Isekai3D SDG, a platform built on reusable environment and character assets and an Unreal Engine plugin for simulating and collecting synthetic 3D gameplay visuals, aimed at research on vision and world models, neural graphics, and game agents.

    Why it matters: The two-stage pipeline separates game simulation from visual capture, so the same stretch of gameplay can repeatedly produce multiple outputs, making it easier to judge the integration cost of the data pipeline.

Nov 2Mon
  1. Tripo38

    Tripo launches desktop 3D Junkyard with model drop-in and text and image search

    Tripo launched the desktop tool 3D Junkyard. Dropped-in models land next to similar ones, and you can search by text or image and fly to matching results. Models in the library can be reused under CC BY 4.0, and Claude Code, Codex and Cursor can also search and contribute models via MCP. The tool is built on Tripo and is a Tripothon S1 project, available at https://zodiac.tripolabs.com/3d-junkyard/.

    Why it matters: The original explains how to search 3D Junkyard and how to connect via MCP, so you can judge whether it fits into your own 3D asset pipeline.