Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,10 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: apache-2.0
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
---
|
| 4 |
+
This repository contains the speaker embedding models for our Zonos-v0.1 [transformer](https://huggingface.co/Zyphra/Zonos-v0.1-transformer) and [hybrid](https://huggingface.co/Zyphra/Zonos-v0.1-hybrid) models.
|
| 5 |
+
|
| 6 |
+
The speaker embedding models are based on the [ResNet293-SimAM-ASP](https://github.com/VoxBlink2/ScriptsForVoxBlink2/tree/main/asv) models from VoxBlink2. We use the pretrain models as we found the finetunes performed worse.
|
| 7 |
+
|
| 8 |
+
The output of the speaker embedding model is then passed through an LDA layer and compressed from 2048 to 1024 dimensions to remove further spurious information about the speaker embedding clip before being fed into the Zonos models.
|
| 9 |
+
|
| 10 |
+
You should not need to manually download these models. The Zonos model loading scripts should load the speaker embedding models automatically.
|