# Standard voice was just the regular text model reading out loud. That's why Live feels dumber — retained record

> Retained reddit record from reddit:ChatGPT.

Canonical: https://highsignal.app/data/records/435e3ded9c521682

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:ChatGPT)  published 2026-11-11 12:06 ingested 2026-11-12 02:37

[original source ↗](https://www.reddit.com/r/ChatGPT/comments/1xtg2ni/standard_voice_was_just_the_regular_text_model/)

## retained text



```
the "dumb receptionist" thing kept coming up in the voice threads so i went and read openai's retirement FAQ. it actually explains it.

standard voice was three things chained together. speech to text, then the normal chatgpt model you type to, then text to speech. slow, but the thing answering you was the full model.

live is a different setup. it's its own voice model that listens and talks at the same time, and when a question gets complicated it hands it off to the text models in the background. so the voice you're talking to isn't the smart part, it's the one passing notes.

after dec 11 everyone gets moved to live. advanced goes too, except live video and screen share keep running on it until live can do those. your old voice chats stay in your history and you can keep going in them by typing.

closest thing to the old chain is doing it by hand. mic button in the text box to dictate, send, then read aloud on the reply. same three steps standard did, you're just the one pressing them.

i build AI Toolbox, on desktop it has ctrl+shift+L to read the latest reply out loud without hunting for the button. free gets 3 a day
```



document fetched 2026-11-12 00:21 · canonical https://reddit.com/r/ChatGPT/comments/1xtg2ni/standard_voice_was_just_the_regular_text_model
