Import new MiniMax and Z-Image workflows

2026-08-16 18:09:01 +00:00
parent f37a363388
commit a7f1188548
4 changed files with 3870 additions and 0 deletions
@@ -0,0 +1,62 @@
+# MiniMax H3 Seamless Chain v2
+
+## Summary
+
+Same-day `MiniMax-H3` multishot workflow from `joeygambino`. This is the full
+graph, not the CORE cut-down version: it adds the LLM prompt lane, remote text
+encoder support, raw-latent continuity, and the speed/VRAM helper switches.
+
+## Model Family
+
+- `minimax`
+
+## Status
+
+- `imported-only`
+
+## Source
+
+- Workflow pack page: <https://huggingface.co/joeygambino/MiniMax-H3-Multishot-Workflow>
+- Workflow file: <https://huggingface.co/joeygambino/MiniMax-H3-Multishot-Workflow/resolve/main/workflows/H3_Seamless_Chain_v2.json?download=true>
+- Original publisher: `joeygambino`
+- Date imported: `2026-08-16`
+
+## Developer notes
+
+- The source README says the full pack adds Motion-Context, RES4LYF, sol-attn,
+  block cache, custom scripts, and GGUF support on top of the base H3 pack.
+- The workflow is built around the `ComfyUI-H3-Multishot` node pack and the
+  same MiniMax-H3 weight family used by the CORE workflow.
+- The prompt-writer lane expects a `qwen3:14b`-class local or remote
+  OpenAI-compatible endpoint if you keep it enabled.
+
+## Our notes
+
+- This is the more interesting same-day MiniMax import if you want the full
+  feature set and are willing to pay for it in complexity.
+- On an `RTX 3070 8 GB`, this is a heavy workflow, but the memory and remote
+  encoder options are the kind of practical extras that make it worth keeping.
+- If you want the simpler starting point, the CORE workflow is still the
+  cleaner local baseline.
+
+## Required custom nodes
+
+- `ComfyUI-H3-Multishot`
+- `ComfyUI-H3-Motion-Context` for the context-pin lane
+- `RES4LYF`
+- `ComfyUI-sol-attn`
+- `comfyui-minimax-h3-blockcache-T8`
+- `ComfyUI-Custom-Scripts`
+- `ComfyUI-GGUF` if you run GGUF weights
+
+## Required model files
+
+- MiniMax-H3 weights from `Comfy-Org/MiniMax-H3`
+- matching MiniMax-H3 text encoder and audio/video VAEs from the same family
+- optional `qwen3:14b`-class prompt-writer model or endpoint if the LLM lane
+  stays enabled
+
+## Notes
+
+- Load `workflow.json` directly in ComfyUI.
+- Best kept as the full-feature reference for MiniMax H3 multishot work.
@@ -0,0 +1,56 @@
+# Z-Image LoRA Facetailer PNG
+
+## Summary
+
+Same-day Z-Image LoRA index with a practical face-detailing workflow PNG. This
+is the useful part of the repo for local ComfyUI work: a loadable graph with
+embedded metadata, not just a plain weight dump.
+
+## Model Family
+
+- `zimage`
+
+## Status
+
+- `imported-only`
+
+## Source
+
+- Model page: <https://huggingface.co/qujincheng/Z-Image-Lora>
+- Workflow asset: <https://huggingface.co/qujincheng/Z-Image-Lora/resolve/main/Workflow/ZI+Facedetailer.png>
+- Original publisher: `qujincheng`
+- Date imported: `2026-08-16`
+
+## Developer notes
+
+- The source repo is a large Z-Image LoRA index, but the workflow PNG is the
+  part worth saving here.
+- The PNG loads directly in ComfyUI because it includes embedded workflow
+  metadata.
+
+## Our notes
+
+- This looks like a practical Z-Image face-cleanup baseline rather than a new
+  model family.
+- The repo is large and opinionated, so treat this as a reusable workflow
+  donor, not a clean official baseline.
+- On an `RTX 3070 8 GB`, this should be a reasonable local test if you already
+  run the smaller Z-Image stack.
+
+## Required custom nodes
+
+- not fully audited from the PNG yet
+- likely `rgthree-comfy`
+- likely a `FaceDetailer` node pack
+
+## Required model files
+
+- not fully audited from the PNG yet
+- likely a Z-Image checkpoint
+- likely a Qwen3-4B text encoder
+- likely a Flux VAE
+
+## Notes
+
+- Load `workflow.png` directly in ComfyUI.
+- Good if you want a lighter face-detailing path around Z-Image.