audio.openai_speech() sound file does not play

[SaKiEQ]

I created a basic sketch, that either plays audio.connecttohost(); or audio.openai_speech();
While it plays sound from host streams just fine, it remains silent when requesting the TTS response.

I get the 'AUDIO_INFO' feedback "audio_info: SSL has been established in 594 ms, free Heap: 134904 bytes", nothing plays.
The API key is fine, when I run a local CURL request with the API key I get a mp3 file as response.

Any pointers what could be wrong ?

[BoltSwith]

I encountered a similar experience in the last two days. I had recently updated the library to Version: 3.0.13x (Updated on: Dec 01.2024).
Connecttohost() plays fine, however, audio.openai_speech() remains silent after returning TRUE.

I managed to get it working somehow after tinkering with the sequence of commands/statements. But this only got rid of the problem temporarily, because, after multiple request/chats/query, the problem seems return until I restart the chip.

Try the following:

1. Create a separate task to initialize and loop through audio.loop();
2. Run the new task on core 1 and assign it a priority of 2 - 3, you can use stack size between 8192 or 10240 bytes.
3. In the new task, call the Audio Lib functions in this sequence

• audio.setConnectionTimeout(6000, 6000);
• audio.openai_speech(OPENAI_API_KEY, "tts-1", YOUR_TEXT_MSG, "shimmer", "mp3", "1");
• audio.setVolume(YOUR_SET_VOLUME);
• audio.setPinout(I2S_BCLK, I2S_LRC, I2S_DOUT);
• while(1){ audio->loop(); vTaskDelay(1); }// loop end

Let me know if this helps.

NB:
I'm using esp-idf version: v5.3.1 ; the esp32s3 SoC (N16R8), Arduino (As a Component) version: v3.1.0-rc1