Minimax H3 Ref2V Quality Issues: Parameter Tuning and Discussion
ShengrenR · reddit · 2026-08-06
The author discusses the quality degradation issues encountered when using Minimax H3's reference image workflows. While standard Image-to-Video (I2V) performs clearly, some reference generations suffer severe degradation resembling wrong VAE usage or JPEG artifacts.
The author found that changing the refimagesize parameter from 'match' to 'max' significantly improves clarity, albeit at a massive cost to generation time. Additionally, using the int8convrot workflow yields crisper I2V gens but sacrifices overall control and reference audio. The author calls on the community to share experiences regarding reference image sizing and trimming.
Related event: MiniMax H3 Local Tests: Fast but Flawed in Quality and Consistency(19 posts)→
More from Multimodal
- Kling 4.0 fully available, shows markedly better motion coherence in fast-action scenes — alifcoder · 2026-10-16
- MJev-Spatial: Single-Pass Classification Beats Autoregressive VLMs on Spatial Reasoning — mindlab-research · 2026-10-16
- H-DDCM: Hadamard Codebooks Make Diffusion-Based Image and Video Compression Scalable — Technion · 2026-10-16
- ComfyUI v0.41.0 ships with Flux3 image model and MiniMax-H3 LightVAE support — Fcking_Chuck · 2026-10-16
- AI-generated short film 'The Dragon Rider' glides cinematically above the clouds — jijo6686 · 2026-10-16
- Realistic candid Midjourney portrait prompt: natural skin, lived-in bedroom light — tisch_eins · 2026-10-16