dev-ansung/tts-flared ? reverse-engineered prompt

Build me a Cloudflare Worker that acts like a simple text to speech API.

I want it to accept a Bearer API key, then let me send text to a POST /v1/audio/speech endpoint and get back an MP3 file with a natural sounding voice. It should feel compatible with OpenAI style TTS, and support voice names like alloy, nova, and also the more specific Edge voice names if I pass them in. Please also add a GET /v1/voices endpoint that returns the available voices as JSON.

Make it work locally with Wrangler and make it easy to deploy to Cloudflare. If the text is long, it should still work by splitting it up and joining the audio back together. Add a couple of simple examples and basic tests, and if you need current docs for Cloudflare Workers or anything related, look them up online.