Dream Beans in India: Google tests AI storytelling outside the command window
Google expanded its experimental app “Dream Beans” to India, making it the first international market the system enters outside the United States, in a bid to reach the world’s second largest internet user base. The app relies on a different concept from usual chatbots, abandoning traditional text commands and proactively generating visual, illustrated stories based on the user’s digital activity and daily data recorded through the company’s services.
The app, available on Android and iOS, uses Google’s personal AI to pull information from linked Google accounts, including Calendar, Gmail, Photos, Google Search, YouTube, and the Gemini app. The company recently added the video-generation model “View” to animate scenes and illustrations within selected stories, and the app is offered in thirteen local languages, including Hindi, Punjabi, Tamil, Telugu, Bengali, alongside English.
Proactive narration redefines user interaction, but usage figures reveal a rapid gap between initial curiosity and sustained retention.Sensor Tower estimates that the app logged roughly 55,000 downloads in the United States since its initial launch in June, with 63 percent of them occurring in September after it became available to all eligible users aged eighteen and older. However, the download rate jumped 84 percent month-over-month in October, reaching only about 5,500 downloads. In a comparison that highlights specialization, the app achieved 29,000 downloads in the first thirty days after opening to the public, versus 7.4 million downloads for Meta’s Muse app, 2.8 million for the ChatGPT app, 362,000 for the Grok app, and 155,000 for the Cloud app over the same period.
Product lead at Google, Gozde Oznur, explains that India was chosen because of a high overlap of user bases across Gmail, Photos and Gemini, together with strong local demand. Yet the experiment runs into clear software limits; although the app accepts inputs in multiple languages, it forces the user to select a single language for story display and currently cannot generate text that mixes two languages, which is a direct obstacle in a linguistic environment that daily relies on automatic code-switching between Hindi and English.
Restricting access to personal accounts and excluding business environments draws a strict line between individual privacy and corporate governance.The app requires access to sensitive personal records to feed the narration, prompting the company to offer options to disconnect any service, erase data, and exclude specific topics such as commercial activities. Deleted information from the original services is automatically removed from the personalization system, with stories blocked from anyone other than the account holder, in an attempt to avoid resurfacing unwanted memories or deceased personas, while the company studies mechanisms to evolve the app from merely delivering stories to suggesting actionable steps.
This shift carries direct implications for decision-makers and tech teams in the Gulf, Egypt and the Arab Levant. In terms of products, the sharp decline in downloads confirms that apps that only provide aesthetic generation face a rapid drain of reader and user interest, compelling local developers to tie generation models to functional workflows and everyday solutions rather than visual showcase. Moreover, Google’s inability to handle mixed-language code-switching gives Arabic platforms a decisive competitive edge, as users in the region constantly blend spoken dialects, Modern Standard Arabic and foreign terms, a technical gap that shows global models still need deep local contextual tuning that centralised algorithms do not readily provide.