Redditor shares optimized ComfyUI workflow for MiniMax H3 Image on a 24GB RTX 3090
Zetalvx-AI-Lab · reddit · 2026-11-14
A Reddit user shared their optimized ComfyUI workflow for MiniMax H3 Image, published on Civitai and GitHub.
- Runs on a single RTX 3090 (24GB)
- Uses an INT8 quantized diffusion model, a Qwen3-VL-32B text encoder, and an fp16 VAE; Euler/Simple sampler, 6 steps, CFG=1
- Resolution adjustable up to 20 MP in the workflow, no upscaling needed
- Generation and image editing share the same model and workflow
- A demo video shows generation and editing examples
More from Multimodal
- Designing a 16-image reference experiment for MiniMax H3 Image room editing — AIdianjiaoshi · 2026-11-14
- 20 gorgeous Western-style paintings (likely AI-generated) going viral — xiaosun86 · 2026-11-14
- Kitten TTS 2 ships three tiny open-source TTS models, smallest just 8MB on CPU — pmttyji · 2026-11-14
- "Different weather": AI-generated ocean liner diner shifts from sun to storm in same scene — nptacek · 2026-11-14
- Redditor shares a 24-second luxury serum ad made entirely with AI — Ashayaki · 2026-11-14
- Reddit user shares cinematic AI short film with strikingly lifelike rainy-night visuals — ZidanStudio · 2026-11-14