Google AI Edge Gallery: In Your Hands but Out of Reach

Last updated on 12 September 2025

Google has released its AI Edge Gallery app. This app is for using AI models that you can download onto your device, for privacy and quick usage. This allows for cool uses like converting audio to text, or just exploring AI. For blind users, though, this app is completely inaccessible, rendering the promise of AI at your hand into a slap in the face.

In this article, I’ll detail what this app is, why blind people might find it useful, how it isn’t accessible, and how this sets a bad example for any Android app developer.

What is AI Edge Gallery

AI Edge Gallery is described this way in the Play Store. A gallery that showcases on-device ML/GenAI use cases and allows people to try and use models locally. The What’s new section has this to say:

• Run Locally, Fully Offline: All processing happens directly on your device.
• Ask Image: Upload images and ask questions about them. Get descriptions, solve problems, or identify objects.
• Audio Scribe: Transcribe an uploaded or recorded audio clip into text or translate it into another language.
• Prompt Lab: Summarize, rewrite, generate code, or use freeform prompts to explore single-turn LLM use cases.
• AI Chat: Engage in multi-turn conversations.

Why This is Important for Blind Users

Here, I’ll go through each of the Gallery’s capabilities, and show why this would be beneficial to blind people:

Run Locally, Fully Offline: All processing happens directly on your device.

Not every blind person lives in areas where there is always Internet access. Furthermore, in locked down situations, it may be important to access these models, particularly image describing ones.

Ask Image: Upload images and ask questions about them. Get descriptions, solve problems, or identify objects.

This one is particularly useful. In fact, getting descriptions is, I’d say, only really important to blind people, since sighted people can already see the image. The fact that the app still isn’t accessible for us but offers this capability shows a lack of testing by the developers at Google, even though there are blind people working there.

Audio Scribe: Transcribe an uploaded or recorded audio clip into text or translate it into another language.

Personally, I get tired of listening to audio sometimes. Being able to just convert it into text and read in Braille or with TalkBack speech would be great for me. It would also be valuable for Deaf-Blind people who obviously cannot hear the audio

Prompt Lab: Summarize, rewrite, generate code, or use freeform prompts to explore single-turn LLM use cases.

Blind people should know about AI. This feature will give them the ability to test things out, and really dig into what AI can and cannot do

AI Chat: Engage in multi-turn conversations.

Can you keep a secret? Your phone can too. And with this, you can talk about it with your phone. Moreover, this is also a great way to see what AI can and cannot do, in the privacy of your own phone.

So, How is the App Not Accessible

When you open the app, TalkBack reports the app name, and nothing else. There is no other accessible item on the screen. You can get an AI description of the screen, but nothing else. There are buttons on the screen, but TalkBack does not see them. This means that we are not able to even get started with using the app.

Why This Sets a Bad Example

Google is a vast company, with tons of money and resources. Google also employs several disabled people, including blind people. Google also has a closed Trusted Tester program of many more blind people.

Google can easily test their apps for accessibility. Even though AI Edge Gallery is a smaller, more experimental app, Google should practice good accessibility just as they try to do with their larger apps, like Google Messages. The fact that they have released an inaccessible app shows that Google has a long way to go in incorporating accessibility across the corporation.

When we talk to other app developers, it helps to have a great app to point to, so that we can say “This is the kind of experience we want.” Unfortunately, Google’s AI Edge Gallery isn’t one of them. If developers model their accessibility off of this app, we’ll have many more inaccessible experiences to come. In fact, Google’s AI Edge gallery is open source at Github, so developers can easily model their interfaces, particularly in AI apps, after this one. In fact, two blind people have filed feedback, Filed feedback on accessibility issues in June, and these issues are not fixed in September, 3 months later. This is a profound disappointment for a company at Google’s scale. How can we expect smaller companies, or indie developers, to make their apps accessible, when Google won’t even do it

Conclusion

On-device AI should be the most private, immediate way for blind users to describe images, transcribe audio, and explore what’s possible—yet today, AI Edge Gallery locks us out at the front door. Accessibility isn’t a “nice to have”; it’s a build requirement, right alongside “works offline.” Google can fix this with basic Android best practices—meaningful labels, proper focus order, Braille/keyboard navigation, and real TalkBack testing with blind users—and keep it fixed with GitHub Continuous Integration accessibility checks . Until that happens, this “gallery” is a museum we can’t enter. If you work at Google, please prioritize a repair; if you’re a developer, let this be a cautionary tale, not a template. AI belongs in everyone’s hands, not just on everyone’s phones.