Scope the download command to the original checkpoint folders

With the diffusers weights merged, the bare `hf download` now fetches both formats, roughly double the bytes. This scopes the documented command to `FL2VA/` and `Ref2VA/` with `--include` patterns, shows the single family variant, and notes that diffusers loading needs no manual download at all.

README.md

@@ -192,12 +192,18 @@ Each checkpoint is distributed as a self\-contained Hugging Face\-style reposito
 └── audio_vae/
 ```
 
-Download the model:
+Download the model. The repository hosts the original checkpoint (`FL2VA/`, `Ref2VA/`) and the diffusers format side by side, so scope the download to what your framework needs:
 
 ```bash
-hf download MiniMaxAI/MiniMax-H3 --local-dir MiniMax-H3
+# Original checkpoint, both task families (SGLang, vLLM, ComfyUI):
+hf download MiniMaxAI/MiniMax-H3 --include "FL2VA/*" "Ref2VA/*" --local-dir MiniMax-H3
+
+# Or a single task family:
+hf download MiniMaxAI/MiniMax-H3 --include "FL2VA/*" --local-dir MiniMax-H3
 ```
 
+diffusers users do not need a manual download: `ModularPipeline.from_pretrained("MiniMaxAI/MiniMax-H3")` fetches exactly the components it needs.
+
 We recommend the following inference frameworks to serve the model:
 
 - [SGLang](https://docs.sglang.io/) \- see [cookbook](https://docs.sglang.io/cookbook/diffusion/MiniMax/MiniMax-H3)