VNCCS 3.2.0 adds Qwen Image 2.1, native alpha sprites, and MiniMax H3 video model for sprite generation
AHEKOT · reddit · 2026-10-05
VNCCS, a ComfyUI pipeline for creating visual novel character sprite sets, released v3.2.0 with major changes: Qwen Image 2.1 is now the main model across Character Creator, pose/clothing workflows and Emotion Studio, and can generate sprites with native alpha transparency that survives clothing edits and SeedVR upscaling. MiniMax H3, a video model, is repurposed for poses, outfit generation and clothes cloning while keeping the first frame as the character image. The Character Creator gets editable tag chips, 40 style presets, 61 species presets, framing options, plus a Character Overhaul LoRA for consistency; a 1–4 MP resolution slider and a unified Control Center round out the update.
More from Multimodal
- AI agent produces a 1:45 video end-to-end on a local PC with Pixio and FFmpeg — tsi_org · 2026-10-05
- Musk promotes Grok Imagine with AI-generated video demo — elonmusk · 2026-10-05
- AI Animated Short 'Altered State' Wins Best Animation at SHAIKE Film Festival — gen_ericai · 2026-10-05
- Spriterrific generates consistent 2D sprite animations without breaking character design — chongdashu · 2026-10-05
- Sailor Moon 'Revenge' Short Made with MiniMax H3 — Certain_Potato_4509 · 2026-10-05
- Seedance 2.5 video samples show a big jump in realism and motion quality — Aiden_Tech_Ai · 2026-10-05