Open image-to-3D on 6GB VRAM: Trellis 2, Hunyuan3D and more all fall short for characters
danihyder47 · reddit · 2026-09-20
A beginner help post that usefully maps the current state of low-VRAM open-source image-to-3D pipelines:
- Tried Qwen Image, Trellis 2 (with textures), Flux2, Pixel 3D, and Hunyuan3D v2.1 in local ComfyUI workflows; single-image textured mesh results all fall short for characters
- Wants multiview-to-3D, but generating an AI turnaround sheet first is a hassle in itself
- Constraints: no paid APIs (meshy AI, tripo3D), only 6GB VRAM, learning Blender in parallel
- Goal: a workable AI-assisted 3D mesh pipeline for game assets; asking for open-source pipeline suggestions
Reflects the real gap of open 3D generation tools for character modeling; the thread may surface usable pipelines.
More from Embodied
- Humanoids Summit kicks off in Seoul in two days, spotlight on humanoid robotics — jonstephens85 · 2026-09-20
- ActionPiece rethinks action tokenization for VLA models, hits 94.8% on LIBERO — burny_tech · 2026-09-20
- GPT-6 Astra for robotics: a curated collection of papers, demos and evals — NielsRogge · 2026-09-20
- Guy wires AI into a Big Mouth Billy Bass fish that gives motivational speeches — scaling01 · 2026-09-20
- Robot Arm Showdown: Jev Completes Cube Task in 27s vs GPT-6 Astra's 1m11s — multiply_matrix · 2026-09-20
- I made a microprocessor: HDL to silicon tapeout explained in new video — jwt0625 · 2026-09-20