update readme
Browse files
README.md
CHANGED
|
@@ -13,6 +13,13 @@ Krea Realtime 14B is distilled from the [Wan 2.1 14B text-to-video model](https:
|
|
| 13 |
|
| 14 |
Inference code can be found [here](https://github.com/krea-ai/realtime-video).
|
| 15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
- Our model is over **10x larger than existing realtime video models**
|
| 17 |
- We introduce **novel techniques for mitigating error accumulation,** including **KV Cache Recomputation** and **KV Cache Attention Bias**
|
| 18 |
- We develop **memory optimizations specific to autoregressive video diffusion models** that facilitate training large autoregressive models
|
|
|
|
| 13 |
|
| 14 |
Inference code can be found [here](https://github.com/krea-ai/realtime-video).
|
| 15 |
|
| 16 |
+
|
| 17 |
+
<video width="100%" controls>
|
| 18 |
+
<source src="https://cdn-uploads.huggingface.co/production/uploads/62a2712903bf94c3ac3ae004/NI1qn109PHVeO_LvBQIr8.mp4" type="video/mp4">
|
| 19 |
+
Your browser does not support the video tag.
|
| 20 |
+
</video>
|
| 21 |
+
|
| 22 |
+
|
| 23 |
- Our model is over **10x larger than existing realtime video models**
|
| 24 |
- We introduce **novel techniques for mitigating error accumulation,** including **KV Cache Recomputation** and **KV Cache Attention Bias**
|
| 25 |
- We develop **memory optimizations specific to autoregressive video diffusion models** that facilitate training large autoregressive models
|