Fixes

Author: GiannisPikoulis

environment.yaml

@@ -32,6 +32,7 @@ dependencies:
     - jiwer==2.3.0
     - piq
     - librosa==0.8.1
+    - phonemizer==3.2.1
     - -e git+https://github.com/CompVis/taming-transformers.git@master#egg=taming-transformers
     - -e git+https://github.com/openai/CLIP.git@main#egg=clip
     - -e .

face_reenactment/README.md

@@ -1,2 +1,3 @@
 ### Preparation
-* Download AffectNet dataset from [here](http://mohammadmahoor.com/affectnet/). Please correct all path variables correspondingly.
+* Download AffectNet dataset from [here](http://mohammadmahoor.com/affectnet/). Please correct all path variables correspondingly.
+* In addition, to use ID loss during CLIP-guided finetuning, you are required to download the pretrained [IR-SE50](https://drive.google.com/file/d/1KW7bjndL3QG3sxBbZxreGHigcCCpsDgn/view) model from [TreB1eN](https://github.com/TreB1eN) and put it in the /pretrained directory.