This demo shows how to do client streaming to generate TTS audio from GPT like response, like ChatGPT.
Rather than creating TTS audio from the entire response and enduring extended latency, ChatGPT returns results in a streaming manner, allowing for TTS audio generation each time a complete sentence is received:

1. Get GPT response token
2. Add the token to a string buffer
3. If token is a sentence end, generate TTS audio from the buffer
4. Repeat 1-3 until GPT response stream is finished
5. Generate TTS audio from the buffer if there is any token left

1. Add below environment variables to your system:

AZURE_OPENAI_ENDPOINT=https://<my-sub>.openai.azure.com/
AZURE_OPENAI_API_KEY=my_aoai_api_key
AZURE_TTS_API_KEY=my_azure_tts_api_key
AZURE_TTS_REGION=eastus_or_other_region

2. Update the query if you want and dotnet run.

private static string query = "Tell me a joke about 100 words.";

dotnet run

[TTS] means there is a TTS call after the token.

Why did the tomato turn red?

[TTS]Because it saw the salad dressing!

[TTS]Now, while that joke might not be 100 words long, it's still worth a chuckle.[TTS] But if you're looking for something a bit more elaborate, here's a longer joke for you:

[TTS]A panda walks into a restaurant and orders a sandwich.[TTS] After eating the sandwich, the panda pulls out a gun, shoots the waiter, and begins to walk out of the restaurant.

[TTS]The restaurant owner, shocked by what just happened[TTS]

The sentence separator is pretty simple, which may do wrong separation for some cases.