Update README.md

README.md

@@ -23,6 +23,8 @@ tags:
 - text-to-speech
 pipeline_tag: text-to-speech
 inference: true
+base_model:
+- suno/bark
 ---
 
 # Bark
@@ -44,6 +46,15 @@ Two checkpoints are released:
 - [**large** (this checkpoint)](https://huggingface.co/suno/bark)
 
 
+## The new weights in this repository
+
+This repository contains .pth and .safetensors weights that are usable within [TTS Webui](https://github.com/rsxdalv/tts-webui).
+The benefits are:
+1. .pth files are pre-filtered using the code from the original repository.
+   These inference-only weights provide size reduction, for example from text_2.pt (Big semantic model) 5.35gb to 1.78gb.
+2. .safetensors files provide compatibility with PyTorch 2.6+, as well as greater safety.
+3. New bf16 safetensors files provide 2x smaller (6x total) weights files, and when used in bf16 mode, double the inference speed.
+
 ## Example
 
 Try out Bark yourself!