MiniMax H3 Image-to-Video AUTO — Batch Sequential, Native Audio, QwenVL Auto-Caption (8-Step Turbo) — v1.0 — Turbo (8-step) — v3 — Dual-LoRA 6-Step

v3 — Speed: dual-LoRA turbo stack (official Comfy-Org 8-step distill + secondary lightx2v LoRA) replaces the single-LoRA v2 setup, running clean at 6 steps (down from 8) — ~3x faster than the original 20-step baseline. Same-seed A/B against v2 showed no quality loss on image or audio.

Length: clips now default to 243 frames (~10.1s @ 24fps), up from 124f (~5.2s) in v2 — full 17k+5 frame grid (124-243) still selectable.

Audio QC hardened this round: adopted a CLAP-based audio/caption similarity check as a pre-filter alongside human-listen review, tested across 3 scene types (campfire, beach, autumn park). Found and root-caused a real scene-content bias in H3's native audio (beach/swimwear-style scenes can trigger music instead of ambient sound) — unrelated to the LoRA/step changes. Neutral scenes (nature, campfire, park) came through clean; example clips for this version were chosen accordingly.