<a target="_blank" rel="ugc" href="https://github.com/NyckM/Bruxosdovfx-SyncTile-MiniMax-H3-v5.4/tree/main">https://github.com/NyckM/Bruxosdovfx-SyncTile-MiniMax-H3-v5.4/tree/main</a><h1 id="the-idea-in-one-sentence">The idea in one sentence</h1>A large or long video is split into overlapping regions (spatial tiles × temporal windows). At every model call all regions are predicted and blended into one canvas, so there is a single denoising trajectory: no tile seams, no chained clips, and VRAM is one region's.<h1 id="install">Install</h1>Easy way (Windows): put <code>bruxosdovfx_EasyInstall_MiniMaxH3.bat</code> in the ComfyUI folder (or the ComfyUI-Easy-Install folder) and double-click it. It finds the folders, installs the custom nodes (SyncTile, Media IO, Latent Upscaler, Easy-Use) and downloads models and LoRAs; anything already present with the same name is skipped, and interrupted downloads resume. Options at the top of the file: text encoder int8/nvfp4, VAE light/full, optional LoRAs, update nodes.Manual:<ol><li><code>git clone https://github.com/NyckM/Bruxosdovfx-SyncTile-MiniMax-H3-v5.4</code> inside <code>ComfyUI/custom_nodes/</code> (or copy the <code>bruxosdovfx-SyncTile</code> folder). Delete <code>ComfyUI-SyncTile-H3</code> if present (old version).</li><li>Restart. The console prints <code>[bruxosdovfx • SyncTile] v5.5 carregado</code>.</li><li>Also install the bruxosdovfx Media IO pack — <a target="_blank" rel="ugc" href="https://github.com/NyckM/Video-Util-ComfYUI">https://github.com/NyckM/Video-Util-ComfYUI</a> (Load Video Pro, Load Image, Save Video Pro — see below). Needs a ComfyUI with the native MiniMax H3 nodes. To write the video file: ffmpeg (on PATH, in <code>FFMPEG_PATH</code>, or <code>pip install imageio-ffmpeg</code>).</li></ol><h1 id="background-replace-v3-lora-(summary)">Background Replace v3 LoRA (summary)</h1>Experimental merge of five LoRAs by other authors for MiniMax H3 (not an original training). Download: <a target="_blank" rel="ugc" href="https://huggingface.co/NyckM/Background_Replace_Minimax_H3/blob/main/Background%20Replace%20v3.safetensors">Background Replace v3.safetensors</a> → <code>models/loras/minimax/</code>. BF16, ~1.91 GB, rank-concatenation merge (no SVD); no metadata (credits live in this README). Components and weights: Facial Realism CloseUp — prithivMLmods (1.0); H3 Character Swap — Akatz Labs (1.0); Realism People — Lovis Odin / fal (1.0); Head Swap v1.0 r32 — Alissonerdx / Alisson Anjos (0.2); VFX Edit v1.0 r128 — Alissonerdx / Alisson Anjos (0.2). Links in the table in the Portuguese section. Base model: <a target="_blank" rel="ugc" href="https://huggingface.co/MiniMaxAI/MiniMax-H3">MiniMax H3 — MiniMaxAI</a>. Load from your LoRA folder at strength 1.0. It mixes FL2VA and REF2VA sources; the name does not guarantee background replacement. The exact Character Swap version is unconfirmed (metadata says v2, deltas resemble v1.5). Structure validated, not yet tested in generation. The source models' licenses apply; credits do not imply the authors endorse the merge.<h1 id="referencias-references">Referências / References / 参考文献</h1>Ideas we combined (no code copied unless stated)Técnica / Technique / 技术Fonte / Source / 来源Onde entra / Where it is used / 用于Fusão de caminhos de difusão em regiões sobrepostas · Fusing overlapping diffusion paths · 重叠区域扩散路径融合Bar-Tal, Yariv, Lipman, Dekel. MultiDiffusion: Fusing Diffusion Paths for Controlled Image Generation. ICML 2023. <a target="_blank" rel="ugc" href="https://arxiv.org/abs/2302.08113">arXiv:2302.08113</a>Base matemática da mistura por região · Math behind the per-region blend · 区域融合的数学基础SyncTile para MiniMax H3 (tiles 1472×896, refino 4K) · SyncTile for MiniMax H3 · MiniMax H3 的 SyncTileTural, A. (2026). <a target="_blank" rel="ugc" href="https://github.com/alaintural/synctile-h3">alaintural/synctile-h3</a> (MIT) · <a target="_blank" rel="ugc" href="https://pixedi.com/lab/minimax-h3-4k-synctile">pixedi.com/lab/minimax-h3-4k-synctile</a> · Zenodo DOI <a target="_blank" rel="ugc" href="https://doi.org/10.5281/zenodo.23196934">10.5281/zenodo.23196934</a>Modo "SyncTile original", tamanhos de tile, refino nos passos finais · tile sizes, final-step refine · 图块尺寸、末步精修Janelas de contexto no tempo com ruído compartilhado · Temporal context windows with shared noise · 共享噪声的时间上下文窗口Qiu et al. FreeNoise: Tuning-Free Longer Video Diffusion via Noise Rescheduling. ICLR 2024. <a target="_blank" rel="ugc" href="https://arxiv.org/abs/2310.15169">arXiv:2310.15169</a> · <a target="_blank" rel="ugc" href="https://github.com/DrakenZA/Comfyui-H3-DrakenNodes">DrakenZA/Comfyui-H3-DrakenNodes</a> (joint context windows for H3)Janelas de 124 frames, ruído único da tela · 124-frame windows, one canvas noise · 124 帧窗口、统一画布噪声Divisão espaço + tempo para upscale do H3 · Spatial + temporal chunking for H3 upscaling · H3 放大的时空分块<a target="_blank" rel="ugc" href="https://github.com/bbaudio-2025/Comfyui-MMH3-UltimateUpscale">bbaudio-2025/Comfyui-MMH3-UltimateUpscale</a>Regiões tile × janela · tile × window regions · 图块 × 窗口区域Encadear clipes (comparação) · Clip chaining (comparison) · 片段拼接（对比）<a target="_blank" rel="ugc" href="https://github.com/mark9009/ComfyUI-H3-LongTake">mark9009/ComfyUI-H3-LongTake</a> e outros packs de vídeo longo · and other long-video packs · 及其他长视频节点包Abordagem que substituímos por uma trajetória só · approach replaced by one trajectory · 被单一轨迹取代的方法Latent upscaler 3D · 3D latent upscaler · 3D 潜空间放大器<a target="_blank" rel="ugc" href="https://github.com/LBH-123-AI/Comfyui_Minimax_h3_latent_Upscaler">LBH-123-AI</a>Entre os estágios 1 e 2 · between stages 1 and 2 · 第一、二阶段之间Modelo fused turbo · Fused turbo model · 融合 turbo 模型<a target="_blank" rel="ugc" href="https://huggingface.co/MATLOWAI/minimax-h3-fused-turbo-int8-convrot">MATLOWAI</a> (ref2va delta + lightx2v turbo + Mystic)Os dois estágios · both stages · 两个阶段VAE leve · Light VAE · 轻量 VAE<a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/MiniMax-H3-experimental">Kijai/MiniMax-H3-experimental</a>Encode/decodeMiniMax H3 + nodes nativos · native nodes · 原生节点<a target="_blank" rel="ugc" href="https://huggingface.co/MiniMaxAI/MiniMax-H3">MiniMaxAI</a> · <a target="_blank" rel="ugc" href="https://huggingface.co/Comfy-Org/MiniMax-H3">Comfy-Org/MiniMax-H3</a> · <a target="_blank" rel="ugc" href="https://github.com/comfyanonymous/ComfyUI">ComfyUI</a>Tudo · everything · 全部Distilação em 4 passos (LoRA opcional) · 4-step distillation (optional LoRA) · 4 步蒸馏（可选 LoRA）Wang, Yuan, Wang et al. PDMD: Projected Distribution Matching Distillation (UC San Diego / ByteDance), 2026. <a target="_blank" rel="ugc" href="https://arxiv.org/abs/2609.35768">arXiv:2609.35768</a> · <a target="_blank" rel="ugc" href="https://huggingface.co/pdmd2026">pdmd2026</a> · conversão ComfyUI <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/MiniMax-H3-experimental/tree/main/loras">Kijai</a>Pilha de LoRA 1 · LoRA stack 1 · LoRA 组 1Nitidez guiada pelo vídeo fonte · Source-guided sharpening · 源视频引导锐化<a target="_blank" rel="ugc" href="https://huggingface.co/Alissonerdx/Minimax-H3-ComfyUI">Alissonerdx / Minimax-H3-ComfyUI</a> (LMS v1.0 r64)Pilha de LoRA 2 · LoRA stack 2 · LoRA 组 2LoRAs da mescla Background Replace v3 · LoRAs in the merge · 合并所用 LoRAprithivMLmods, Akatz Labs, Lovis Odin / fal, Alissonerdx / Alisson Anjos (ver acima · see above · 见上文)LoRA opcional · optional LoRA · 可选 LoRA

Files with SHA256: bc08147eface0b0747be48c030395d2adc1e31779a01c9ffa9125a50579594b1