OpenAI · 2024–26 · speech-to-speech
Not a read-the-script TTS in the old sense. GPT-4o Audio and the later Realtime models (including GPT-Realtime-2 in 2026) generate speech as part of a conversation. The Sky clips are OpenAI’s 2023 announcement demos.
End-to-end spoken dialogue, not a vocoder bolted onto a chatbot.
Vendor / paper demo
Sky (ChatGPT Advanced Voice, 2023) · Official Sky story demo from OpenAI’s 2023 voice announcement CDN.
OpenAI announcement demo: Sky reading a story. Published with ‘ChatGPT can now see, hear, and speak’ (2023). Different script; labelled as vendor demo.
Vendor / paper demo
Sky (ChatGPT Advanced Voice, 2023) · Official Sky info demo from OpenAI’s 2023 voice announcement CDN.
OpenAI announcement demo: Sky in an informational register. Same 2023 CDN set as the story clip.