v0.2.0-beta.1 — initial beta release (mcode internal beta group)
v0.2.0-beta.1 — initial beta release
🧪 Status: beta test build for the mcode internal beta group.
This is the standalone single-plugin mirror of the official PR
MiniMax-AI/MiniMax-Code-Plugins#15.
The official PR is open and waiting for review. Until that merge lands, this repo is the
public test channel for the mcode internal beta group.
Install
git clone https://github.com/antianqi/comfyui-studio.git
# point mcode at the repo root (the plugin is the whole repo, not a sub-folder)Full instructions (3 install options + the 6-scenario test matrix + 3 feedback channels) are
in BETA.md.
What's inside
- 4 Skills (
comfyui-studio/comfyui-workflow/comfyui-character/comfyui-drama) - 6 preset workflows covering scenarios 1–6
- 1 生图 —
workflows/selfie-text-to-image.json - 2 模仿 —
workflows/selfie-mimicry.json - 3 改图 —
workflows/flux2-klein-image-edit.json - 4 融合 —
workflows/flux2-klein-image-edit-dual.json - 5 首帧 —
workflows/drama-first-frame.json - 6 出片 —
workflows/drama-image-to-video.json
- 1 生图 —
- A dependency-free stdio MCP server (
server.mjs, ~200 lines, Node stdlib) - A dependency-free Python CLI (
skills/comfyui-workflow/scripts/submit_workflow.py) - Full docs (
README.md+BETA.md+ 3SKILL.md+examples/+docs/)
How to give feedback
- GitHub issues (preferred for reproducible bugs):
https://github.com/antianqi/comfyui-studio/issues - Feishu mcode internal beta group: just @ me in the channel where the announcement
landed. Best for "it works but the UX is awkward" / "the docs say X but the workflow does Y". - PR comments on the official PR:
MiniMax-AI/MiniMax-Code-Plugins#15
Best for structural / design feedback.
Known issues
- Scenario 2 (
selfie-mimicry.json) needsComfyUI-LLaMA-CPP+comfyui_controlnet_aux+
rgthree-comfy+ComfyUI-Impact-Pack; submissions will fail withmissing_node_type
on installs without those. The full list is inskills/comfyui-character/SKILL.md. - All LoRA
lora_namefields are generic placeholders (your_face_lora.safetensors,
your_style_lora.safetensors,character_a_lora.safetensors, etc.). You must edit
them to point at your own LoRA files before submission. The Plugin does not bundle any
private LoRAs. - All checkpoint / VAE / CLIP / ControlNet fields are the reference install's real filenames.
On a different install, edit theckpt_name/clip_name1/vae_name/control_net_name
fields to match your files.
License
Apache-2.0.