http-openai-tts

An HTTP microservice using OpenAI to generate TTS.

More documentation added soon!

Caution

Requests with the max input length can cost you between 3 and 5 cents of OpenAI credits. If config.json's "needKey" is turned off and/or your input length is not monitored bad actors could run up an expensive bill for you or use up all of your credits extremely fast.

When testing it took me not even 3 minutes (OpenAI only allows 50 RPM) to send enough requests (~135) to spend $5 of credits. That's about $90-100 if unsupervised for 1 hour.

Models (Optional, Defaults to tts-1):

• tts-1
• tts-1-hd

Voices (Optional, Defaults to echo):

• alloy
• echo
• fable
• onyx
• nova
• shimmer

Response formats (Optional, Defaults to mp3):

• mp3
• opus
• aac
• flac
• wav
• pcm

Speed (Optional, Defaults to 1):

Select a value between 0.25 and 4.0.

Text:

The maximum text is 4096 characters.

Example requests:

• GET /tts?model=tts-1-hd&voice=alloy&format=mp3&speed=1&text=Hello%20World
• GET /raw?model=tts-1&voice=echo&format=wav&speed=1.3&text=Foo%20Bar

If "allowHosting" is set to false /tts automatically returns the /raw output.

Error response format for /tts & /raw:

{
	"status": "Error description",
	"message": "Information about the response.",
	"error": "Provides either an actual go error, a premade error or an OpenAI API reponse."
}

Response for /tts:

{
	"status": "OK",
	"message": "Information about the response.",
	"link": "Contains a link to the hosted file."
}

Response for /raw:

Just returns the audio file content.

Response of /ping:

{
	"status": "OK",
	"message": "Pong!"
}

Admin Controls (adding & removing keys)

• POST /admin/add?name=Alvin&key=cm8f4a250816
• POST /admin/remove?name=Theodore

Response for /admin/add:

{
	"status": "OK",
	"message": "Created the key Alvin successfully.",
	"key": "cm8f4a250816"
}

Response for /admin/remove:

{
	"status": "OK",
	"message": "Removed the key Theodore successfully."
}