The API suggests that any language supported by Whisper can be queried for, but requesting either Indonesian or Spanish (both on the list) I recieved garbled nonsense output in both cases. I checked both for tts-1 and tts-1-hd, the result was the same.
An example:
response = openai.audio.speech.create(
model=“tts-1-hd”,
voice=“coral”,
input=“Mereka bekerja keras serta penuh dedikasi”,
)
response.stream_to_file(speech_file_path)
The output just sounds like human blabbering, but it does not sound like any language, certainly not Indonesian.
The phonemes don’t even remotely match up- it’s not as though an English speaker was doing a poor Indonesian accent. It’s completely different output.
The docs say in multiple places that this is supported.