| --- |
| license: other |
| license_name: ltx-2-community-license-agreement |
| license_link: LICENSE |
| base_model: Lightricks/LTX-2.3 |
| tags: |
| - video-generation |
| - lora |
| - camera-control |
| - world-model |
| - social-interaction |
| --- |
| |
| # HelloWorld V1 |
|
|
| The social-interaction LoRA for **HelloWorld: Enabling Socially Interactive |
| Characters in Video World Models**. |
|
|
| 📄 Paper: [arXiv:2608.05070](https://arxiv.org/abs/2608.05070) · [HF paper page](https://huggingface.co/papers/2608.05070) |
|
|
| - **Base model:** LTX-2.3-22B distilled ([`Lightricks/LTX-2.3`](https://huggingface.co/Lightricks/LTX-2.3)) |
| - **File:** `helloworld_lora_v1.safetensors` (rank-32 LoRA) |
| - **What it does:** conditions LTX-2.3 on a camera-warp video so the generated |
| clip follows a user-supplied camera trajectory, while an F-key interaction |
| window makes the on-screen character respond toward the camera — turning to |
| the viewer, waving, nodding, or speaking a short greeting. |
|
|
| ## Example videos |
|
|
| [`example_videos.zip`](example_videos.zip) (86 MB) contains the generated |
| results of the **seven official examples** shipped with the GitHub repo |
| (same recipes as [`assets/examples/`](https://github.com/AlayaLab/HelloWorld/tree/main/assets/examples), |
| previewed as GIFs in the repo README) — full-quality 1280×704 @ 24 fps MP4s |
| with the WASD / arrow / `F` keyboard overlay and the co-generated audio track |
| (the character speaks in `garden_pond`, `park_yoga`, `anime_figures`, |
| `moor_crows`, and `lakeside_hikers`; `bear_mascot` and `shore_skeleton` are |
| silent by design). |
|
|
| ## Usage |
|
|
| Inference code, environment setup, and seven ready-to-reproduce examples: |
| **https://github.com/AlayaLab/HelloWorld** |
|
|
| ```bash |
| git clone https://github.com/AlayaLab/HelloWorld |
| # put this LoRA at HelloWorld/checkpoints/helloworld_lora_v1.safetensors |
| cd HelloWorld/inference |
| INPUT=examples_batch.json GPU=0 bash run_batch.sh |
| ``` |
|
|
| ## License |
|
|
| This LoRA is fine-tuned from **LTX-2.3** and trained on data synthesized by |
| LTX-2.3, which makes it a *Derivative of LTX-2* as defined in the |
| **LTX-2 Community License Agreement**. Accordingly, it is distributed |
| exclusively under that agreement — the complete license text is included in |
| this repository as [`LICENSE`](LICENSE). In particular: |
|
|
| - Your use of this model (and anything you build on it) is subject to the |
| use-based restrictions in **Attachment A** of the agreement. |
| - **Commercial entities with annual revenues of at least $10,000,000** must |
| obtain a separate paid commercial license from Lightricks — see |
| <https://ltx.io/model/licensing>. |
| - If you redistribute this model or further derivatives of it, you must do so |
| under the same agreement and include a complete copy of the license. |
|
|
| LTX-2 is © Lightricks Ltd. This repository is not affiliated with or endorsed |
| by Lightricks; all rights in the base model remain with the Licensor. |
|
|
| ## Citation |
|
|
| ```bibtex |
| @article{ouyang2026helloworld, |
| title = {HelloWorld: Enabling Socially Interactive Characters in Video World Models}, |
| author = {Ouyang, Liangyang and Liu, Ruicong and Chu, Xuangeng and Zhang, Kaipeng and Sato, Yoichi}, |
| journal = {arXiv preprint arXiv:2608.05070}, |
| year = {2026} |
| } |
| ``` |
|
|
| ## Contact |
|
|
| oyly@iis.u-tokyo.ac.jp / liangyang.ouyang@shanda.com |
|
|
