Controlled Test Compares 8 Popular MiniMax H3 Speed LoRAs With Same Seed and Prompt
Better-Interview-793 · reddit · 2026-09-22
A Reddit user ran a controlled comparison of popular MiniMax H3 LoRAs using the same Will Smith reference image, seed, prompt, and settings.
Setup: 2144×1216 resolution, Euler sampler, Simple scheduler, LoRA strength 1.0; the benchmark is a 5-second restaurant scene with continuous hand, utensil, food, facial, and lip-sync motion.
Contestants (8-step): PDD 8-Step, Silver Dareties, Turbo Multistep V4 Step600 Delta, LightX2V Hybrid 4-to-8 Step, Turbo V4 Step600 EMA; (4-6 step): FL2V Turbo 4-Step, REF2V Turbo 4-Step, Taomate 3-Step.
Takeaways: several are technically R2V LoRAs, but the base FL2VA model already handles reference conditioning and the author prefers it; running more steps than designed barely changes results and sometimes helps audio. The post includes the full structured prompt (subjectdefinitions / retentionanalysis / detaileddescription / soundscape) for reproduction.
Related event: User Tests All MiniMax H3 Speedup LoRAs, Finds Best Combo(2 posts)→
More from Multimodal
- Ditto launches swap-anything video tool to replace any video's main character — yoobinray · 2026-11-12
- Fable 5.2 impresses with pixel art gaussian splats, posts go viral — Flomerboy · 2026-11-12
- Seedream 5.5 nails a consistent character sticker set at just $0.02 per image — azed_ai · 2026-11-12
- Ant Group's open-source Ming-3.0-tts-0.5B tops Open TTS Leaderboard with 0.88 WER — NielsRogge · 2026-11-12
- Viggle Remake recreated in ComfyUI: swap any character in your videos with MiniMax H3 — TheLocalLab · 2026-11-12
- Pika Integrates Seedream 5.5: Split Images Into 100+ Layers for Precise Multi-Turn Edits — aziz4ai · 2026-11-12