Simple temporal upscaling workflow fixes motion blur in MiniMax H3 videos
Inner-Reflections · reddit · 2026-10-01
Reddit user Inner-Reflections shares a proof-of-concept ComfyUI workflow for simple temporal upscaling to fix motion blur in MiniMax H3 (and other) video models.
Core idea: instead of pixel upscaling + rediffusion (limited gains) or selectively upscaling frames like MAINodes, 'derope' the whole scene — temporally upscale 2x and rediffuse over everything. Benefits: better small-detail motion consistency, tunable denoise, good results from a plain 2x upscale. Cost: double the diffusion, so pair with low-step LoRAs and sol attention (which helps more the longer the video).
Practical tips:
- Workflow stays close to base ComfyUI, using Kijai's nodes and VHS Suite plus one audio-slowdown node
- Base video at 0.5 MP / 20 steps to find a seed, then 2x temporal upscale to 1 MP
- Tune upscale denoise to length: 0.45 is a good start, 0.3–0.6 acceptable; too high speeds up and blurs the video
- Disable sol attention on the base if backgrounds morph
- Go 3x/4x if action is still too fast
Workflow published on Civitai.
More from Multimodal
- Google Arts & Culture launches nom nom: a Neural Cellular Automata world simulated by Gemini — zzznah · 2026-10-01
- Stanford's UniEvo-VL Self-Distillation Lifts Qwen-image GenEval From 0.747 to 0.808 — stanfordnlp · 2026-10-01
- Early user feedback: Suno v6 disappoints on music generation quality — hq4ai · 2026-10-01
- Looped DiT: 260M model beats 6.5x larger rival on T2I benchmarks with 4.9x less compute — arankomatsuzaki · 2026-10-01
- Runway teases mysterious Project Continuum with early preview at AI Summit — c_valenzuelab · 2026-10-01
- Viggle Turbo v0.3 for Qwen Image: cleaner 6-step output, new 9-step mode fixes small text — init-5 · 2026-10-01