Adding custom f5-tts model "Error while deserializing header: HeaderTooLarge"

[S-T-K]

I tried to add this f5 tts model to alltalk https://huggingface.co/sinhprous/F5TTS-stabilized-LJSpeech
Followed the instructions here #471

But whenever I try to load the model, I get this error:

Error loading model: Error while deserializing header: HeaderTooLarge

Edit: After concerting the .pt file to .safetensors (using this approach: https://github.com/DiffusionDalmation/pt_to_safetensors_converter_notebook/blob/main/pt_to_safetensors_converter.ipynb) I get a new error

Error loading model: 500: Error loading model: Error(s) in loading state_dict for CFM: Unexpected key(s) in state_dict: "duration_predictor.conv_1.bias", "duration_predictor.conv_1.weight", "duration_predictor.conv_2.bias", "duration_predictor.conv_2.weight", "duration_predictor.norm_1.beta", "duration_predictor.norm_1.gamma", "duration_predictor.norm_2.beta", "duration_predictor.norm_2.gamma", "duration_predictor.proj.bias", "duration_predictor.proj.weight", "duration_predictor.text_embed.weight".

Edit2: couldn't even get it to work with the original f5 repo installation, so I guess it's an issue with the model itself

[S-T-K]

I just noticed that the author of the model didn't just train a new model, but also adjusted a lot in the base f5 repo code, so it can't work in Alltalk anyway.