Model card: agent-made games
Browse files
README.md
CHANGED
|
@@ -17,6 +17,8 @@ AnyGameRenderer takes a crude gameplay recording (the "programmer art" agents pr
|
|
| 17 |
|
| 18 |
This repository holds the **LoRA** (`anygamerenderer_lora.safetensors`, 2.4 GB) that turns [MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3) Ref2VA into AnyGameRenderer. Paper: https://arxiv.org/pdf/2610.31466 路 Project page: https://alaya-lab.github.io/AnyGameRenderer/ 路 Inference code, the caption builder, and examples: https://github.com/AlayaLab/AnyGameRenderer.
|
| 19 |
|
|
|
|
|
|
|
| 20 |
- 馃幃 **Generalizes across genres:** one checkpoint renders tower defense, auto-battlers, top-down survivors, turn-based battles, kart racing, crowd runners, arena survival, co-op games, and shooters.
|
| 21 |
- 馃幆 **Keeps entity correspondence:** group counts stay correct, entities appear and disappear on the right frame, and small far-away entities are not lost.
|
| 22 |
- 馃摴 **Its input is plain video:** gray blockouts, colored primitives, semantic maps, wireframes, and coarse mockups all work, with no per-object labels or engine hooks.
|
|
|
|
| 17 |
|
| 18 |
This repository holds the **LoRA** (`anygamerenderer_lora.safetensors`, 2.4 GB) that turns [MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3) Ref2VA into AnyGameRenderer. Paper: https://arxiv.org/pdf/2610.31466 路 Project page: https://alaya-lab.github.io/AnyGameRenderer/ 路 Inference code, the caption builder, and examples: https://github.com/AlayaLab/AnyGameRenderer.
|
| 19 |
|
| 20 |
+
Trained on paired crude/finished recordings of agent-made games and engine scenes.
|
| 21 |
+
|
| 22 |
- 馃幃 **Generalizes across genres:** one checkpoint renders tower defense, auto-battlers, top-down survivors, turn-based battles, kart racing, crowd runners, arena survival, co-op games, and shooters.
|
| 23 |
- 馃幆 **Keeps entity correspondence:** group counts stay correct, entities appear and disappear on the right frame, and small far-away entities are not lost.
|
| 24 |
- 馃摴 **Its input is plain video:** gray blockouts, colored primitives, semantic maps, wireframes, and coarse mockups all work, with no per-object labels or engine hooks.
|