File size: 3,234 Bytes
f79f4b2
2a96a9e
 
 
72b2a93
 
 
 
 
 
 
f79f4b2
72b2a93
 
 
 
 
 
a7c84ed
 
2a96a9e
72b2a93
 
 
 
 
 
23131a3
 
 
 
 
 
 
 
 
 
 
72b2a93
 
 
 
 
 
 
 
 
 
 
 
2a96a9e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a7c84ed
 
 
 
 
 
 
 
 
 
 
72b2a93
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
---
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