text-to-3d

The author shares their complete AI-powered production pipeline for turning "The Little Prince" into an explorable 3D game using Opus 5.5 + Tripo + Blender + Unity, along with reusable prompts. Since the tutorial went public, it has racked up nearly 3 million views across the web, and the code will be open-sourced next week.

Google's Gemini 4 Argon reportedly used multiple self-critiquing AI agents to autonomously create a photorealistic pagoda temple complex in Blender from a single prompt, with no human intervention. The demonstration highlights advances in multi-agent, end-to-end 3D content generation.

AI can now generate working 3D game objects from text descriptions entirely through code, with a 100-object benchmark finding Astra most reliable while Opus 5.5 produces visually better results.

SpaceFlow 提出了一种免训练的局部可控 3D 生成流程，用户可通过几何基元为不同区域指定控制强度，并分别路由文本或图像提示以合成外观，从而在几何保真与生成自由度之间取得平衡。

This paper audits configuration fragility in text-to-3D evaluation by holding 300 generated scenes fixed while varying render and caption settings, finding that configuration variance often exceeds between-generator variance and that VLM-based evaluators frequently shift their reported winner without any change in the underlying scenes.

Sorceress 3D Studio offers a one-stop AI pipeline that converts text descriptions into game-ready 3D animations, supporting auto-rigging and various motion generation, and can be directly exported to game engines.

Block3D accelerates text-to-3D generation by using block-wise diffusion with confidence-guided correction to reduce inference time while preserving geometric fidelity, achieving a 5.15x speedup.

PixSDS identifies VAE-induced pixel drift in latent score distillation sampling and proposes a gradient repair method that decodes latent SDS lookahead steps to guide pixel-space optimization, reducing artifacts in text-to-3D generation.

WorldClaw is an agentic coarse-to-fine framework that generates large-scale, explorable 3D open worlds from text prompts, using planning agents and terrain-conditioned composition to produce coherent global structure and editable instance-level assets.

A review of the newly released Ling 3.0 flash model, which can generate 3D worlds from a single text prompt.

CGGS is a text-to-3D framework that improves geometric consistency and quality in ego-centric 3D scene generation using a multi-stage approach with consistency-augmented loss, layout decoration, and geometric refinement via entropy-based depth loss.

SimWorlds is a multi-agent framework that generates dynamic, editable 4D scenes from natural language, using Blender-specific procedural knowledge and a planner-coder-reviewer workflow, outperforming prior baselines.

Arbor introduces explicit geometric control for 3D asset generation by using constraint meshes (hull, avoidance, touch regions) to condition latent generation, improving spatial constraint adherence without sacrificing object quality.

Fable 5 has been released, capable of generating games, cities, or recreating classics with just one sentence, demonstrating the powerful ability of text-to-3D content.

This paper identifies and addresses 'latent sink traps' in text-to-3D generative models where they become insensitive to text prompts, proposing a framework that decouples geometric representation from linguistic sensitivity to enable robust text-based 3D shape editing of out-of-distribution shapes.

HY-World 2.0 is a multi-modal world model framework that generates high-fidelity 3D Gaussian Splatting scenes from text, images, and videos through specialized modules for panorama generation, trajectory planning, and scene composition, achieving state-of-the-art performance among open-source approaches.

微软提出统一的结构化潜在表示（SLAT），结合 rectified flow transformers 实现高质量、多格式（Radiance Fields、3D Gaussians、网格）的 3D 资产生成，支持文本/图像条件及局部编辑，在 500K 3D 数据集上训练的 2B 参数模型显著超越现有方法。

OpenAI introduces Point-E, a system for generating 3D point clouds from text prompts in 1-2 minutes on a single GPU by combining text-to-image and image-to-3D diffusion models. The method achieves significant speedup over prior methods while releasing pre-trained models and code.