Doing manual inference with XTTS

[jez9999]

I'm trying to do manual inference with XTTS so I can change the various parameters. I've already managed to get it to output stuff with tts.tts_to_file(), but with model.inference() I have to first call model.get_conditioning_latents() and that requires me to provide an audio file: audio_path=["reference.wav"]. What is this audio file? I didn't have to provide it with tts.tts_to_file(), I just provided a speaker name. Is there a way for me to use the speaker name with model.inference()?

[eginhard]

As written on Discord: https://github.com/idiap/coqui-ai-TTS/blob/f329072df200bf1473f881dd5afe854660d1d9fe/TTS/tts/models/xtts.py#L400