OpenMOSS releases Apache 2.0 MOSS-Image: 5B DiT with CJK text rendering and JSON layout control
crystal_alpine · reddit · 2026-10-20
OpenMOSS released MOSS-Image, now supported in ComfyUI under Apache 2.0. Architecture: a 5B DiT with a Qwen3.5-9B text encoder and FLUX.2 VAE, in Base and SFT (curated-data fine-tuned) variants. Highlights: legible English and Chinese text rendering, structured JSON prompts with bounding boxes to control layout (useful for posters and ads), and good out-of-box aesthetics. Suggested settings: 1024×1024, 50 steps, CFG 3, shift 5; no step distillation, so it's slow. BF16 weights total 27GB (10GB DiT + 17GB encoder), but community INT8/INT4 quants run on 8GB cards. Workflow on HuggingFace at Comfy-Org/MOSS-Image.
More from Multimodal
- AI pipeline translates Schuller's German-only mechanics lectures, preserving his voice with LaTeX notes — TOEwithCurt · 2026-10-20
- Anthropic Dev Shares the Prompts Behind His Opus 5.5 Code-Rendered Films — dfeinition · 2026-10-20
- Designer shares 16-color pairing guide and full Midjourney prompt for editorial still life — tisch_eins · 2026-10-20
- Flux 3 Max I2V test: a full prompt for generating a photoreal horror forest opening shot — umesh_ai · 2026-10-20
- Stylized animation made with Flux 3 Max on fal, then upscaled — BLUECOW009 · 2026-10-20
- Matt Shumer marvels at AI's rapid 3D generation leap in under a year — mattshumer_ · 2026-10-20