OpenAI’s API changelog says prompt-to-voice became available to API customers on October 16. A developer can describe the character of a voice in text, generate a reusable custom voice and use it in GPT-Live sessions. The voice creation step is free, while generating speech is billed at the regular GPT-Live session rates.
This is particularly interesting for Wake Up. Imagine a short spoken version of the three front-page stories in a consistent, natural-sounding voice while making coffee. That would be a new consumption option rather than a replacement for the daily email or website. It could also be useful for audio narration of a McGruff investigation or an internal software walkthrough.
The sensible first test is a 90-second recording of one already-published article, comparing pronunciation, pacing and actual API cost with the voices available on Apple devices. Keep it private and separate from the MiniMac publishing workflow until the result is good enough to use.
What to do: Generate a private 90-second Wake Up audio sample using prompt-to-voice and measure the real API cost.