Update README.md

README.md

@@ -13,92 +13,89 @@ tags:
   - minimax-h3
 ---
 
-# LightVAE: Towards Compact and Efficient Video Autoencoders
+# ⚡ LightVAE
 
-LightVAE accelerates pretrained video autoencoders through temporal low-rank compression and block pruning while preserving their latent interfaces.
+**Towards Compact and Efficient Video Autoencoders**
 
-This repository provides checkpoints for **Wan2.1, Wan2.2, and MiniMax-H3**, including exports for our ComfyUI custom node.
+[💻 Code](https://github.com/ModelTC/LightVAE) · [🧩 ComfyUI](https://github.com/ModelTC/LightVAE/tree/main/comfyui) · [⚡ LightX2V](https://github.com/ModelTC/LightX2V)
 
-**Paper:** *LightVAE: Towards Compact and Efficient Video Autoencoders*  
-**Code and examples:** [ModelTC/LightVAE](https://github.com/ModelTC/LightVAE)
+LightVAE accelerates video autoencoders while preserving their original latent interfaces. We provide compressed components for **Wan2.1, Wan2.2 and MiniMax-H3**, with native ComfyUI integration for H3.
 
 ## 📦 Checkpoints
 
-| Backbone | Variant | Python checkpoint | ComfyUI checkpoint |
-| :-- | :-- | :-- | :-- |
-| Wan2.1 | Pro decoder | [Download](./lightvae-pro-wan21-decoder.safetensors) | [Download](./lightvae-pro-wan21-decoder-comfy.safetensors) |
-| Wan2.1 | Lite decoder | [Download](./lightvae-lite-wan21-decoder.safetensors) | [Download](./lightvae-lite-wan21-decoder-comfy.safetensors) |
-| Wan2.2 | Pro decoder | [Download](./lightvae-pro-wan22-decoder.safetensors) | [Download](./lightvae-pro-wan22-decoder-comfy.safetensors) |
-| Wan2.2 | Lite decoder | [Download](./lightvae-lite-wan22-decoder.safetensors) | [Download](./lightvae-lite-wan22-decoder-comfy.safetensors) |
-| MiniMax-H3 | Lite decoder | [Download](./lightvae-lite-h3-decoder.safetensors) | [Download](./lightvae-lite-h3-decoder-comfy.safetensors) |
-| MiniMax-H3 | Lite encoder | [Download](./lightvae-lite-h3-encoder.safetensors) | [Download](./lightvae-lite-h3-encoder-comfy.safetensors) |
+| Backbone | Component | Download |
+| :-- | :-- | :-- |
+| Wan2.1 | Pro decoder | [safetensors](./lightvae-pro-wan21-decoder.safetensors) |
+| Wan2.1 | Lite decoder | [safetensors](./lightvae-lite-wan21-decoder.safetensors) |
+| Wan2.2 | Pro decoder | [safetensors](./lightvae-pro-wan22-decoder.safetensors) |
+| Wan2.2 | Lite decoder | [safetensors](./lightvae-lite-wan22-decoder.safetensors) |
+| MiniMax-H3 | Lite decoder | [safetensors](./lightvae-lite-h3-decoder.safetensors) |
+| MiniMax-H3 | Lite encoder | [safetensors](./lightvae-lite-h3-encoder.safetensors) |
 
-**Pro** retains the original decoder depth for higher reconstruction fidelity. **Lite** applies additional pruning for faster inference.
+**Pro** uses temporal low-rank compression while retaining the original decoder depth. **Lite** additionally prunes blocks for Wan; H3 Lite uses block pruning.
 
-Wan models retain their original encoders. MiniMax-H3 can use either its original encoder or the provided Lite encoder.
+Wan models reuse their original encoders. H3 supports independently replacing the encoder and decoder.
 
-> **Wan2.2 model selection:** Wan2.2 **14B** uses the **Wan2.1 VAE**; Wan2.2 **5B TI2V** uses the **Wan2.2 VAE**. Choose the decoder accordingly.
+> **Choose the correct Wan VAE:** Wan2.2 **14B** uses the **Wan2.1 VAE**; Wan2.2 **5B TI2V** uses the **Wan2.2 VAE**.
 
-## ⚡ Use with ComfyUI
+## 🧩 ComfyUI · MiniMax-H3
 
-> **Important:** The `-comfy.safetensors` checkpoints require our **Load LightVAE (TLR)** custom node. Do **not** load them with ComfyUI's standard **Load VAE / VAELoader** node.
+The **Load H3 LightVAE** node reuses ComfyUI’s native H3 VAE. No LightVAE package installation, TLR or custom Triton kernels are needed. The current node is **H3-only**.
 
-### 1. Install the node and download weights
+### Install
 
-Taking **MiniMax-H3** as an example, run these commands in the **same Python environment used by ComfyUI**:
+Update ComfyUI to a version with native MiniMax-H3 support. Run with the Python environment used by ComfyUI:
 
 ```bash
 cd /path/to/ComfyUI
 
 git clone https://github.com/ModelTC/LightVAE.git LightVAE
-python -m pip install ./LightVAE huggingface_hub
 python LightVAE/comfyui/install.py --comfy-root .
 
+python -m pip install huggingface_hub
 hf download lightx2v/LightVAE \
-  lightvae-lite-h3-decoder-comfy.safetensors \
+  lightvae-lite-h3-decoder.safetensors \
+  lightvae-lite-h3-encoder.safetensors \
   --local-dir models/vae/LightVAE
 ```
 
-Restart ComfyUI, then add a **Load LightVAE (TLR)** node.
+Already cloned? Run `git -C LightVAE pull`, then rerun the installer.
 
-### 2. H3 FL2V: replace the decoder
+The installer creates `custom_nodes/ComfyUI-LightVAE/` automatically. **Do not clone the entire repository directly into `custom_nodes`.**
 
-In **Load LightVAE (TLR)**, select:
+The original H3 video VAE is required. If missing:
 
-- `decoder`: `lightvae-lite-h3-decoder-comfy.safetensors`
-- `encoder`: `none`
-- Leave `precision`, `backend`, and `device` set to `auto`.
-
-Connect its **VAE output** to your existing **VAE Decode** node.
+```bash
+hf download Comfy-Org/MiniMax-H3 \
+  vae/minimax_h3_video_vae_fp16.safetensors \
+  --local-dir models
+```
 
-Keep the original VAE connections for first/last-frame encoding unchanged. For this decoder-only connection, `base_vae` can remain disconnected.
+### Connect
 
-The same decoder-only replacement also works for **REF2V**.
+Restart ComfyUI and search for **Load H3 LightVAE**.
 
-### 3. H3 REF2V: optionally replace the encoder too
+Load the **original H3 video VAE** with the standard **Load VAE** node, then connect it to LightVAE’s **`base_vae`** input.
 
-To also accelerate reference-image/video encoding, download the Lite encoder:
+- **FL2V:** select the Lite decoder and leave `encoder = none`. Keep original first/last-frame encoding unchanged; connect LightVAE to **VAE Decode**.
+- **REF2V:** connect LightVAE to both reference encoding and **VAE Decode**. Select either or both Lite components.
+- **`none`** uses the corresponding original component. Encoder-only replacement is supported with `decoder = none`.
 
-```bash
-hf download lightx2v/LightVAE \
-  lightvae-lite-h3-encoder-comfy.safetensors \
-  --local-dir models/vae/LightVAE
-```
+**Keep the audio VAE unchanged.** Load Lite components through **Load H3 LightVAE**, not the standard Load VAE node.
 
-Restart ComfyUI and select both H3 Lite files in **Load LightVAE (TLR)**:
+The standard H3 files above and previous `*-h3-*-comfy.safetensors` exports are both accepted.
 
-- `decoder`: `lightvae-lite-h3-decoder-comfy.safetensors`
-- `encoder`: `lightvae-lite-h3-encoder-comfy.safetensors`
+### Workflows
 
-Connect its **VAE output** to both the **reference-encoding node** and **VAE Decode**. No original video VAE is needed for these connections.
+Download a JSON and drag it into ComfyUI:
 
-**Keep the H3 audio VAE and its connections unchanged in all cases.**
+[FL2V](https://github.com/ModelTC/LightVAE/blob/main/comfyui/workflows/h3_fl2v_lightvae.json) · [Image references](https://github.com/ModelTC/LightVAE/blob/main/comfyui/workflows/h3_ref_image_lightvae.json) · [Video reference](https://github.com/ModelTC/LightVAE/blob/main/comfyui/workflows/h3_ref_video_lightvae.json)
 
-See the [ComfyUI guide and example workflows](https://github.com/ModelTC/LightVAE/tree/main/comfyui) for more details and Wan examples.
+See the [workflow guide](https://github.com/ModelTC/LightVAE/tree/main/comfyui/workflows) for required base models and input preparation.
 
-## 🐍 Quick start with Python
+## 🐍 Python · Video Reconstruction
 
-Use the standard checkpoints **without** the `-comfy` suffix:
+Example: **original Wan2.1 encoder + LightVAE-Lite decoder**.
 
 ```bash
 git clone https://github.com/ModelTC/LightVAE.git
@@ -108,14 +105,9 @@ python -m pip install -r requirements.txt
 hf download lightx2v/LightVAE \
   lightvae-lite-wan21-decoder.safetensors \
   --local-dir weights/LightVAE
-```
 
-For example, reconstruct a video with the Wan2.1 Lite decoder and original encoder:
-
-```bash
 hf download Wan-AI/Wan2.1-T2V-1.3B \
-  Wan2.1_VAE.pth \
-  --local-dir weights/Wan2.1
+  Wan2.1_VAE.pth --local-dir weights/Wan2.1
 
 python infer.py \
   --model wan21 --variant lite \
@@ -125,8 +117,8 @@ python infer.py \
   --output outputs/wan21_lite_recon.mp4
 ```
 
-See the [GitHub repository](https://github.com/ModelTC/LightVAE) for Wan2.2 and MiniMax-H3 examples, benchmarks, and visual comparisons.
+More examples, benchmarks and visual comparisons are available in [LightVAE](https://github.com/ModelTC/LightVAE). For video generation with LightVAE, see [LightX2V](https://github.com/ModelTC/LightX2V).
 
-## ⭐ Support Us
+## ⭐ Support
 
-If you find LightVAE useful, please consider starring [LightVAE](https://github.com/ModelTC/LightVAE) and [LightX2V](https://github.com/ModelTC/LightX2V). Your support helps us keep improving efficient video generation!
+If you find this useful, please star [LightVAE](https://github.com/ModelTC/LightVAE) and [LightX2V](https://github.com/ModelTC/LightX2V)!