Upload README.md with huggingface_hub

README.md

@@ -1,5 +1,5 @@
 ---
-license: apache-2.0
+license: mit
 tags:
   - world-model
   - video-generation
@@ -95,26 +95,26 @@ In the current inference script, `action_speed_list` is used as the relative dur
 
 ### Camera Action Commands
 
-DreamX-World-5B uses `WASD` for camera translation and `IJKL` for view rotation:
+DreamX-World-5B uses `WASD` for camera translation and `IJKL` for camera rotation:
 
-| Action | Description |
-|--------|-------------|
-| `w` | Move forward |
-| `s` | Move backward |
+| Action | Camera Control |
+|--------|----------------|
+| `w` | Push in |
+| `s` | Pull out |
 | `a` | Move left |
 | `d` | Move right |
-| `i` | Look up |
-| `k` | Look down |
-| `j` | Look left |
-| `l` | Look right |
+| `i` | Tilt up |
+| `k` | Tilt down |
+| `j` | Pan left |
+| `l` | Pan right |
 
 Actions can be composed in one string:
 
-- `wi`: move forward while looking up
-- `wk`: move forward while looking down
-- `wj`: move forward while looking left
-- `wl`: move forward while looking right
-- `dj`: move right while looking left
+- `wi`: push in while tilting up
+- `wk`: push in while tilting down
+- `wj`: push in while panning left
+- `wl`: push in while panning right
+- `dj`: move right while panning left
 
 ### Run Inference
 
@@ -178,27 +178,19 @@ Because the default causal block size is 3 latent frames, `num_output_frames` sh
 | **Causal Block Size** | 3 latent frames per block by default |
 | **VAE** | Wan2.2 VAE, temporal compression 4x, spatial compression 16x |
 | **Text Encoder** | UMT5-XXL |
-| **Denoising Steps** | `[1000, 750, 500, 250]` in the provided AR config |
 | **Precision** | BFloat16 |
 
-## Limitations
+## WeChat Group
 
-- Very long rollouts can still accumulate visual artifacts, camera-control errors, or color/style drift.
-- Camera actions are approximate controls, not guaranteed physical camera trajectories.
-- The provided inference script resizes inputs to 704 x 1280 and uses fixed normalized camera intrinsics.
-- `action_speed_list` controls relative segment duration in the provided script; it does not directly change the per-frame camera velocity.
-- Inference requires a CUDA GPU and substantial VRAM. Longer generations require more time and memory.
-- Generated videos may reflect biases or artifacts from the training data and base model.
+Join our WeChat group for discussion:
 
-## Intended Use
-
-DreamX-World-5B is intended for research and creative applications in interactive world simulation, camera-controllable video generation, long-horizon image-to-video generation, game-like environment exploration, and promptable world modeling.
-
-The model should not be used to create deceptive, harmful, illegal, or rights-violating content, or to represent generated videos as real-world footage.
+<div align="center">
+  <img src="assets/wechat_group_qrcode.png" width="300" alt="WeChat Group QR Code">
+</div>
 
 ## License
 
-This model is released under the [Apache License 2.0](LICENSE).
+This model is released under the [MIT License](LICENSE).
 
 ## Citation