Google takes Dreambeans to India with 13 languages and a privacy question attached

What did Google launch in India?

Google expanded Dreambeans to India on Monday, November 2, making it the app's first market outside the United States,

TechCrunch reported. The app runs on Android and iOS and now supports 13 local languages alongside English, including Hindi, Punjabi, Tamil, Telugu and Bengali.

The timing is tight. Google opened Dreambeans to all eligible US users aged 18 and older less than two months before this launch, after introducing it in June as a Google Labs experiment. India is therefore the first place where the app has to work for users whose Google accounts look nothing like a US early-adopter account.

The product itself is unusual. Dreambeans does not wait for a prompt. It reads information from connected Google services, including Calendar, Gemini, Gmail, Photos, Search and YouTube, and assembles a small set of illustrated stories about the user's interests and day. Google's September announcement described each dreambean as a mix of personalised topics and distilled information from the apps a user selects, and noted that connecting Google Photos lets the app incorporate visuals of the user and people close to them.

How does the app decide what to show you?

The help material on

Google Labs' Dreambeans page explains the mechanism and the name. Google pairs dream with beans: the dreaming is background processing that finds meaning across a user's connected apps, and the beans are the fresh set of stories waiting each morning. Google says the app needs at least one connected app to function, takes up to a day to build the first set of stories, and applies tuning corrections to future drops rather than the current one.

Users can steer it with a thumbs-down button or a tuning control, bookmark stories, and share them. Google says featuring the user and their loved ones requires Google Photos face grouping to be enabled. The company also warns that Dreambeans is experimental and may surface an irrelevant story or an inaccurate visual.

TechCrunch reported that Google recently added a feature that uses its Veo video-generation model to animate images in select stories. That turns a static illustrated feed into something closer to short generated video, which is the more interesting product direction and the more expensive one to run.

How big is Dreambeans actually?

Not big. Sensor Tower estimates shared with TechCrunch put US downloads at nearly 55,000 since the June debut, with about 63 percent of them arriving in September when Google opened the app to all eligible US adults. October downloads fell 84 percent month over month to roughly 5,500.

The first-30-day comparison TechCrunch published is the number that should worry Google, because it measures pull rather than cumulative totals.

Those figures compare apps with different jobs. Dreambeans generates stories instead of answering arbitrary queries, so a smaller addressable audience is expected. But a consumer AI product that peaks on launch and then drops 84 percent is a product that has not found a habit, and adding a second country does not fix that by itself.

Why India, and what limits it there?

Google's stated reason is demand and overlap. Product manager Gozde Oznur said during a press briefing that India stood out for its high overlap of users across Gemini, Gmail and Google Photos, and that Indian users had been asking for access. That is a distribution argument: India is the second-largest internet market, and Google already owns the accounts the app needs.

The limitation is language switching. Oznur said users can select only one language at a time for generated stories, and that the app cannot currently generate stories that mix languages, even though it can draw on information across Google services in multiple languages. In a market where Hindi-English mixing is the default register for a large share of urban users, single-language output is a real constraint rather than a footnote.

There is also a product-category question. Dreambeans is a proactive feed, which means it competes for the same attention as every other feed on the phone while requiring users to hand over access to mail, photos and calendar to be useful at all. That is a higher activation cost than a chatbot, and the US download curve suggests most people who try it do not come back.

What is the privacy trade?

Dreambeans runs on Google's Personal Intelligence technology, which is the same layer the company has been pushing into other products. We covered how that layer interacts with advertising in

our report on Gmail and Photos data being usable for ads once connected to Personal Intelligence, and the same underlying question applies here: how much of your personal context are you handing to a feed generator to get a nicer morning read?

Google's answers, per TechCrunch, are controls rather than architecture. Users choose which services to connect and can disconnect at any time. They can delete their Dreambeans data from inside the app, and Oznur said information deleted from the underlying connected service is also removed from Dreambeans' personalisation system. Users can tell the app to avoid certain subjects, including business-related information. The app is limited to personal Gmail accounts and not enterprise users, and generated stories are visible only to the account holder.

The sharpest admission is about grief. Oznur acknowledged that Dreambeans could generate stories involving people who have died, and said users can tell the app to stop showing certain subjects, comparing the controls to Google Photos features that let users limit which memories appear. A system that proactively surfaces memories needs an answer for the ones people do not want to see, and a subject filter is a partial answer.

Google's control set is genuinely more granular than most consumer AI apps. It is still opt-in at the app level rather than the data level, and the default path to a working product is to connect several of the most sensitive services Google holds.

What would make this work?

Two things, and neither is a language count. The first is a reason to open the app on day eight, not just day one. Google's own framing is that it is exploring ways to make Dreambeans more actionable, moving from stories and recommendations toward helping users act on them. That is the right direction, because a recommendation feed without an action is a nicer version of a notification.

The second is whether the generated stories are good enough to be worth the data. The failure mode is not offensive output, it is bland output: a story about a trip you already took, illustrated with a photo you have already seen. Google says the tuning feature fixes this over time, but tuning corrections only apply to future drops, so the first week is the product.

For context on how fast the underlying model layer is moving, our coverage of

Gemini 4 Argon reaching general availability and of Gemini's Personal Intelligence reaching Android Auto shows the same pattern: Google is pushing context-fed AI into surfaces where it already has the data, and the open question is whether users want the output.

Judged as news, this is a product expansion with a real limitation and a weak usage curve. Judged as strategy, it is Google testing whether proactive generation can carry a consumer app at all outside its home market. The download numbers in India over the next two months will answer that more honestly than any launch post.