POST API for infer

[ban1989ban]

Created API for TTS:

Usage:
#(A) Get JSON-Encoded Base64
curl -X POST "http://localhost:8000/tts/"
-H "Content-Type: application/json"
-d '{"text": "Hello, how can I help you?", "response_type": "json"}'

#(B) Save audio to file
curl -X POST "http://localhost:8000/tts/"
-H "Content-Type: application/json"
-d '{"text": "Hello, how can I help you?", "response_type": "file"}'
-o output.wav

#(C) Stream Audio in real time
curl -X POST "http://localhost:8000/tts/"
-H "Content-Type: application/json"
-d '{"text": "Hello, how can I help you?", "response_type": "stream"}'
--output - | aplay

[SWivid]

Hi @banank1989 ,

Maybe you could help provide some infos about the streaming performance compared with https://github.com/SWivid/F5-TTS/blob/main/src/f5_tts/socket_server.py

TIA~

[ban1989ban]

@SWivid Havent run benchmark on both but I guess both serves different purpose. Code i wrote expose API while socket file do not. Also socket sends the audio in one go once it is generated while API code does actual streaming (stream when the chunk is ready)

[SWivid]

Also socket sends the audio in one go once it is generated while API code does actual streaming (stream when the chunk is ready)

From code, seems the same? Both wait for infer_batch_process to finish, just that here you return directly while socket version provide a play_steps_in_s for extra control.
And from what you want in #763 , streaming performance is the key point we should consider first and yes was the original purpose of people propose socket script.

[goranskular]

But this is for a different purpose. It would be nice to have this API in it. @ban1989ban thanks for the code.