OpenAI said on Oct. 15 U.S. time in a post on its official X account that it has launched GPT-Live-1.1, a new voice model for ChatGPT, Codex, and its persistent agent dots. The company said the update brings more natural accents, adds voices in French, German, Hindi, Mexican Spanish, Korean, and Japanese, and reduces how often the system repeats “checking” while performing search tasks.
OpenAI said the rollout to users starts immediately. The API version is listed as “coming soon.”
Plan tiers use different voice model versions
According to ChatGPT release notes, OpenAI is also introducing GPT-Live-1.1 Quip, a smaller version of the model. In ChatGPT voice, Live mode uses different versions depending on the user’s subscription tier. Free and Go users get the mini Quip version, while Plus and higher tiers use GPT-Live-1.1.
OpenAI also said Live will use a wider range of shorter phrases during search to indicate that it is looking up information.
Usage limits by plan
OpenAI’s ChatGPT voice documentation lists general conversation usage on a rolling 24-hour basis:
- Free: GPT-Live-1.1 Quip, limited usage, quota may change
- Go: GPT-Live-1.1 Quip, 3 hours
- Plus: GPT-Live-1.1, 3 hours
- Pro ($100 per month): GPT-Live-1.1, 15 hours
- Pro ($200 per month): GPT-Live-1.1, unlimited, subject to abuse guardrails
For business users, Business Standard includes 3 hours and Business Premium includes 15 hours. After that, usage costs 1.25 credits per minute. Enterprise plans billed in U.S. dollars are priced at $0.05 per minute.
New voices include Japanese and Korean
The documentation lists the new Live voices as Aube and Rivage for French, Sonna and Miko for German, Leher and Neer for Hindi, Himari and Rindō for Japanese, Haetsal and Hanul for Korean, and Tinta and Arrullo for Mexican Spanish. OpenAI also added one English voice called Bubbie.
The company said available voices may vary by region and account, and users can switch them under Settings → Voice. The newly added options do not include Chinese, though users can still ask ChatGPT to speak other languages during a conversation.
Current feature limits
The documentation also says Live mode does not currently support video or screen sharing. Mobile users who need those features can use the previous-generation Advanced voice mode instead.
ChatGPT’s default personality settings also do not currently apply to Live, though users can still ask the assistant to adjust tone and speaking speed during a conversation.
GPT-Live-1 API opened last month
GPT-Live is OpenAI’s voice model that can listen and speak at the same time. It can respond in real time when users interrupt and can hand off harder reasoning tasks to a background model. OpenAI only opened GPT-Live-1 to the API in September, pricing the front-end voice layer at $0.05 per minute.
Version 1.1 is now live first in ChatGPT, Codex, and dots. OpenAI has not disclosed API launch timing or pricing for GPT-Live-1.1.