ComfyUI node H3 Clip Refiner fixes style drift when stitching video clips
embryo10 · reddit · 2026-10-12
The open-source ComfyUI node H3 Clip Refiner tackles quality degradation when concatenating video clips without a camera cut. Placed between SamplerCustomAdvanced and the H3 Motion Context Save Latent node, it works at the latent level in two halves:
- Statistical (always on): per-channel affine matching (mean + std) plus high-band gain matching, fixing drift in luminance, contrast, saturation and the "texture ratchet" — but pure arithmetic can't restore lost micro-detail.
- Generative (optional): connect model + positive/negative and it runs a short low-denoise (default 0.4, 3 steps) refine resample, transferring only its high band (detail) onto the original low band (structure), avoiding background shift.
Usage: run the first clip through with referencelatent unconnected (measure only), then feed it as reference for all later clips; set contextlength to match Motion Context (5/22/39/56 are whole latent steps); strength can safely overshoot to 1.5–2.0 since overlap frames stay raw, preventing seam flash.
More from Multimodal
- Procedural MetaHumans in three.js With an Easy Skinning Approach — jasonkneen · 2026-10-13
- Long Shot Studio 0.8.2 adds 64-shot chaining and auto ComfyUI crash recovery — R34vspec · 2026-10-13
- Surreal AI image series "Sometimes, we need fire" goes viral — gizakdag · 2026-10-13
- Antonio Banderas to star in AI-powered feature 'Cortés' from Utopai Studios, now shooting in Paris — lmoroney · 2026-10-13
- Minimax H3 Refmod workflow: character-consistent video from photos plus a voice sample, no LoRA training — AlsterwasserHH · 2026-10-13
- An indie game ad made entirely with Google Flow and Suno — ahmettnq · 2026-10-13