OpenAI is upgrading ChatGPT Voice with a new real-time voice model designed to make spoken conversations sound more natural across languages, including Hindi. The rollout introduces GPT-Live-1.1 and GPT-Live-1.1 Quip, a smaller version of the model, alongside two new Hindi voice options named Neer and Leher. The update also adds voices for French, German, Japanese, Korean and Mexican Spanish, as well as a new English voice called Bubbie.
The release is particularly relevant to India, where OpenAI is expanding its focus on voice-based AI and multilingual experiences. The improvements target pronunciation, conversational flow and the short spoken updates ChatGPT gives while searching for information. Rather than being a Hindi-only product, GPT-Live-1.1 is a broader voice-model upgrade that includes Hindi among several supported languages. The rollout is gradual, meaning users may see the new options at different times depending on their account and region.
Key Takeaways
- New voice model: OpenAI is rolling out GPT-Live-1.1 and GPT-Live-1.1 Quip in ChatGPT Voice.
- Hindi voices: Users may see two new options, Neer and Leher.
- More languages: New voice options are also being introduced for French, German, Japanese, Korean and Mexican Spanish.
- English addition: Bubbie is the new English voice option.
- Natural speech: OpenAI says the update improves accent naturalness and makes voice interactions feel less repetitive.
- Search updates: ChatGPT Voice will use a wider variety of short phrases to indicate that it is searching.
- Gradual rollout: Availability can vary by account and region.
- How to access it: Users can check Settings, then Voice, to see which options are available.
What Is GPT-Live-1.1?
GPT-Live-1.1 is OpenAI’s updated real-time voice model for spoken conversations with ChatGPT. It is designed to handle the dynamics of live speech, including pauses, interruptions and changes in speaking pace.
Traditional text-based AI generally follows a simple pattern: a user submits a message, the system processes it and then returns an answer. Real-time voice interaction is more complicated because the system must manage a conversation as it unfolds.
It needs to determine when a speaker has finished a thought, whether an interruption requires a response and when it should continue listening instead of speaking. These decisions affect how natural and responsive an AI assistant feels.
OpenAI’s GPT-Live family is designed for this more continuous form of interaction. GPT-Live-1.1 builds on that approach, with improvements to accent naturalness and the way ChatGPT communicates during searches.
The company is also introducing GPT-Live-1.1 Quip, a smaller model that succeeds GPT-Live-1 mini. The full GPT-Live-1.1 model is the default for Plus and higher-tier users, including use in Codex and Dots, while GPT-Live-1.1 Quip is the default for other eligible users, according to OpenAI’s system card.
The update is therefore not simply a collection of new voices. It also involves changes to the underlying models used to power spoken interactions.
Why the Hindi Upgrade Matters for India
India is an important market for OpenAI, and voice-based interaction can make AI tools easier to use for people who prefer speaking to typing.
A user might want to ask ChatGPT a question while commuting, cooking, working or carrying out another activity where typing a long prompt is inconvenient. Someone who is more comfortable speaking Hindi than writing English may also find a voice interface easier to approach.
OpenAI Applications CTO Vijaye Raji has described the company’s focus on Hindi voice capabilities as part of its effort to make ChatGPT feel more natural for Indian users. The company is adapting its product priorities to local usage patterns, with voice and multilingual support among the areas receiving attention.
The distinction between understanding Hindi and speaking it naturally is important. An AI assistant may already be able to respond in Hindi, but the quality of a spoken conversation also depends on pronunciation, rhythm, intonation and the way sentences flow.
A response can be grammatically correct but still sound unnatural if the pronunciation or emphasis feels awkward. More suitable voice options can make an interaction more comfortable, particularly during longer conversations.
However, the update does not mean that every Indian language or regional accent is now fully supported. Hindi is one part of a broader multilingual rollout, and users may still encounter pronunciation or recognition problems with regional expressions, names and mixed-language speech.
Meet Neer and Leher, ChatGPT’s New Hindi Voices
The two new Hindi voice options are named Neer and Leher. They join the expanded voice lineup introduced with GPT-Live-1.1.
For users, the most immediate benefit is choice. Different people prefer different voice characteristics, and having additional options can make the assistant feel more personal.
The new voices are also relevant to users who regularly switch between Hindi and English. In everyday Indian conversations, people frequently combine languages, use English technical terms within Hindi sentences or express familiar ideas through regional phrasing.
A natural-sounding Hindi voice could make explanations, brainstorming sessions and everyday questions more comfortable to follow. It may also be useful for people who want to practise speaking or listening in a language.
Still, a new voice is not the same as a guarantee of perfect language understanding. Voice selection affects how responses sound, while the quality of the answer also depends on the model’s understanding of the request and the information it uses.
OpenAI has not promised that Neer and Leher will handle every regional accent or mixed-language sentence flawlessly. Their practical value will become clearer as users try them in real conversations.
GPT-Live-1.1 Adds 13 New Voices Across Languages
The Hindi additions are part of a wider expansion of ChatGPT Voice. The rollout introduces 13 new voice options across six language varieties, plus Bubbie in English.
| Language or variety | New voice options |
|---|---|
| English | Bubbie |
| French | Two new voices |
| German | Two new voices |
| Hindi | Neer and Leher |
| Japanese | Himari and Rindō |
| Korean | Haetsal and Hanul |
| Mexican Spanish | Tinta and Arrullo |
The expansion reflects a broader effort to make spoken AI more adaptable to users’ languages and preferences.
Voice options can influence how accessible a product feels. Some users may prefer a particular accent or speaking style because it is easier to understand, while others may choose a voice that feels more conversational.
The options shown in ChatGPT can vary by region and account, so the full list may not be available to every user immediately.
The wider language support also does not mean every language receives identical model capabilities. Speech recognition, pronunciation and conversational performance can differ across languages and individual use cases.
More Natural Accents and Smoother Conversations
OpenAI says GPT-Live-1.1 brings more natural accents across languages. This improvement targets an important part of voice interaction: making speech sound appropriate to the language being used.
Natural delivery involves more than pronouncing individual words correctly. It includes sentence rhythm, emphasis, pauses and intonation. When those elements are handled well, a spoken answer can be easier to understand and less tiring to follow.
The improvements could be useful in a range of everyday situations, from asking
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
