Support new OpenAI text-to-speech API

[zzhang1]

This is a proposal to please consider supporting the new GPT-4o mini TTS API as documented here: https://platform.openai.com/docs/guides/text-to-speech.

The key benefits are more natural sounding voices and configurable control over things like emotion, tone, accents, and so on. Examples are here: https://www.openai.fm/ and in addition to being cheaper than the older API, these make for a much more natural sounding listening experience.

[yoni2528]

Agree on that, it will be awesome!

[gilgonore]

Agree on that too!!

[sammydeprez]

You can already do that.
Just change the model via the conversation config

from podcastfy.client import generate_podcast

audio_file = generate_podcast(
    urls=["https://en.wikipedia.org/wiki/Artificial_intelligence"],
    llm_model_name="gpt-4-turbo",
    api_key_label="OPENAI_API_KEY"
    conversation_config={
        "text_to_speech": {
            "openai": {
                "model": "gpt-4o-mini-tts"
            }
        }
    }
)