OpenAI Japan Pairs ChatGPT Voice With Obachaaan

OpenAI Japan released a ChatGPT Voice promotional video in September 2026 starring Obachaaan, an idol group from Osaka. In the US video released in July, older women talk with voice AI; in the Japanese version, ChatGPT joins the banter in the Kansai dialect. The video shows users interrupting an AI response with a question or asking for search and interpreting while continuing the conversation.

OpenAI announced GPT-Live, the voice model behind ChatGPT Voice, on July 8. Its “full duplex” design lets the system listen and speak at the same time, enabling conversation much like an ordinary phone call. Users can ask a question or rephrase something while it responds, and a brief pause to gather one’s thoughts does not immediately end the conversation. In the Japanese video, a cast member asks ChatGPT to “say it more simply,” and it explains the feature by comparing it to a phone call. Interruptions and pauses themselves demonstrate how the system works.

The videos use different everyday subjects in each country. The US version starts with knitting, while the Japanese version begins with advice on a gift for a grandchild; Obachaaan, speaking frankly in Kansai dialect, takes the role of the “grandma” in the US video. GPT-Live can hand questions requiring search or complex reasoning to another model and bring the result back into the conversation, allowing users to keep talking while it processes. The US video asks about the history of acoustics, San Francisco traffic and weather. In the Japanese video, users discuss which flowers to bring to a flower exhibition, travel from Sakai to Ashiyagawa Station, Kobe weather and what to wear outside. The conversation grows more specific through follow-up questions rather than by listing every condition at the outset.

The interpreting scenes also differ. In the US video, a user negotiates the price of an antique book in French; in the Japanese version, ChatGPT explains mixed juice to a foreign guest and interprets the joke, “If you drink this, you’ll become beautiful like me.” A request to “tell this person in English” shows how ChatGPT can mediate an exchange without the user typing out a sentence.

On October 15, OpenAI announced the phased rollout of GPT-Live-1.1 and its smaller version, GPT-Live-1.1 Quip, for ChatGPT Voice. The update improves accents across languages and adds voices in French, German, Hindi, Japanese, Korean and Mexican Spanish, as well as “Bubbie” in English. It also adds short phrases that let users know when a search is underway. Available voices may vary by region and account. Promotional videos for the UK, Germany and France also appeared that month, after the Japanese video had already been released.

The UK video features three older men using voice conversation to discuss a family trip to Casablanca, speaking Spanish and solving a crossword. In a scene suggesting the shoot has ended, one asks, “Could you pick up my grandchild for me?” and gets the reply, “Once you’ve learned to make tea.” The German video takes place in a hair salon, where three women discuss a cycling trip, hairstyles and a botany presentation; it also demonstrates improved natural German pronunciation and accents. The French video features older women Monique and Sophie and a younger man, Hakim, and was reported alongside the rollout of the new French voices “Aube” and “Rivage.” A response to being corrected about someone’s age—“I knew that, I was just testing”—also appears in the US version; in the French version, others tease, “You always say that when you get it wrong.” Across the country versions, the consultation and interpreting scenarios recur, while the cast, everyday topics and conversational tone vary.

OpenAI Japan released a ChatGPT Voice promotional video in September 2026 starring Obachaaan, an idol group from Osaka. In the US video released in July, older women talk with voice AI; in the Japanese version, ChatGPT joins the banter in the Kansai dialect. The video shows users interrupting an AI response with a question or asking for search and interpreting while continuing the conversation.

OpenAI announced GPT-Live, the voice model behind ChatGPT Voice, on July 8. Its “full duplex” design lets the system listen and speak at the same time, enabling conversation much like an ordinary phone call. Users can ask a question or rephrase something while it responds, and a brief pause to gather one’s thoughts does not immediately end the conversation. In the Japanese video, a cast member asks ChatGPT to “say it more simply,” and it explains the feature by comparing it to a phone call. Interruptions and pauses themselves demonstrate how the system works.

The videos use different everyday subjects in each country. The US version starts with knitting, while the Japanese version begins with advice on a gift for a grandchild; Obachaaan, speaking frankly in Kansai dialect, takes the role of the “grandma” in the US video. GPT-Live can hand questions requiring search or complex reasoning to another model and bring the result back into the conversation, allowing users to keep talking while it processes. The US video asks about the history of acoustics, San Francisco traffic and weather. In the Japanese video, users discuss which flowers to bring to a flower exhibition, travel from Sakai to Ashiyagawa Station, Kobe weather and what to wear outside. The conversation grows more specific through follow-up questions rather than by listing every condition at the outset.

The interpreting scenes also differ. In the US video, a user negotiates the price of an antique book in French; in the Japanese version, ChatGPT explains mixed juice to a foreign guest and interprets the joke, “If you drink this, you’ll become beautiful like me.” A request to “tell this person in English” shows how ChatGPT can mediate an exchange without the user typing out a sentence.

On October 15, OpenAI announced the phased rollout of GPT-Live-1.1 and its smaller version, GPT-Live-1.1 Quip, for ChatGPT Voice. The update improves accents across languages and adds voices in French, German, Hindi, Japanese, Korean and Mexican Spanish, as well as “Bubbie” in English. It also adds short phrases that let users know when a search is underway. Available voices may vary by region and account. Promotional videos for the UK, Germany and France also appeared that month, after the Japanese video had already been released.

The UK video features three older men using voice conversation to discuss a family trip to Casablanca, speaking Spanish and solving a crossword. In a scene suggesting the shoot has ended, one asks, “Could you pick up my grandchild for me?” and gets the reply, “Once you’ve learned to make tea.” The German video takes place in a hair salon, where three women discuss a cycling trip, hairstyles and a botany presentation; it also demonstrates improved natural German pronunciation and accents. The French video features older women Monique and Sophie and a younger man, Hakim, and was reported alongside the rollout of the new French voices “Aube” and “Rivage.” A response to being corrected about someone’s age—“I knew that, I was just testing”—also appears in the US version; in the French version, others tease, “You always say that when you get it wrong.” Across the country versions, the consultation and interpreting scenarios recur, while the cast, everyday topics and conversational tone vary.