Update: Since my ElevenLabs credit ran out, I'm now using the SpeechT5_TTS model.