It’s not uncommon for blind Android users to witness visually impaired apps primarily or exclusively launching on iOS, with new features consistently arriving on iOS before Android. Among these was Microsoft’s Seeing AI, originally exclusive to iOS users. However, after years, the app has finally made its debut on Android. While some may have deemed its arrival belated and without much to offer in terms of competition, Seeing AI managed to establish itself as one of the compelling Android apps catering to the visually impaired from its very first day on the platform.
From the moment I began testing the app, it impressed me. Here are the reasons that make me excited about Seeing AI and why I consider it a notable development.
Table of Contents
Document Reading:
While numerous Android apps can recognize text quite well, Seeing AI displayed exceptional text recognition capabilities during my testing with only 2 pages. The app guides users on parts of the page that are not visible, facilitating the capture of the entire page. Once all edges are visible, the app prompts users to hold steady and automatically captures the photo, resulting in accurate outcomes in my testing.
What distinguishes Seeing AI is its capacity to respond to queries about the recognized text. I inquired about the page’s topic and the frequency of the letter ‘r’ in the text, and i received responses. Additionally, you can include another page in the scan or rescan the same page. When sharing, you have the choice to share either the extracted text or the original image.
The app also has a short text reading feature where it reads text instantly as it appears in the camera view.
Detailed Image Description:
To access detailed image recognition, you have the option to either take a photo for the app to describe its content using the “scene preview” feature or directly share an image to the app for description. Initially, you receive a basic short description, but tapping on “more info” provides a more detailed description powered by AI. The recognition performance is fairly good; results arrive quickly, offering details and descriptions of objects and people. However, in comparison to the GPT model used in Bing Chat, Seeing AI’s descriptions sometimes lag behind. While it performed well with basic photos, it struggled, for example, to recognize a Lego construction that both Bard and Bing Chat identified before.
An intriguing addition is the “explore” option, allowing users to move their finger on the screen and hear the objects beneath, providing a means for visually impaired individuals to imagine the scene depicted in the image.
As with all current image recognition services, vigilance is required for occasional inaccuracies and hallucinations. This became evident when I took a random picture in a poorly lit room, where the AI generated several fake details about the items and the room. Note that you can also share the actual image with others, which is a useful feature.
It’s important to note the absence of the ability to ask follow-up questions about the recognized images content for now .
Product Identification:
A crucial feature for some blind users is using barcodes scanning to identify products and packages, and Seeing AI excels in this aspect as well. It effectively recognized various items, such as a popular instant noodle brand, a pasta brand, and mayonnaise, among others, through barcode scanning accessed via the “product” channel. While searching for a barcode, the app uses beeps to indicate proximity to the code, which is particularly helpful as many packages don’t have easily locatable codes. The beeps increase in intensity and speed as you get closer to the code. Once detected, the code is automatically captured, and relevant information is displayed.
Unlike Lookout, Seeing AI doesn’t require the confusing manual selection of product database that should be downloaded to achieve effective recognition specific to products available in your country.
Straightforward Interface:
While most of the individual features themselves may not be unique, Seeing AI displays all these functionalities in an accessible, simple manner. The app’s interface is straightforward, ensuring easy navigation and exploration of the available features without complicating things for users. Additionally, users have the option to customize the order of features or “channels” within the app. This customization extends to specifying shortcuts that appear when long-pressing the app’s name on the home screen or in the apps list, allowing swift access to desired features without the need to open the app first.
Help is There to Explain Everything:
Seeing AI offers extensive guidance for nearly every feature, including video demonstrations. This approach eliminates the need for new users to search for written or audio tutorials externally. Instead, they can rely on the in-app tutorials to seamlessly navigate through its various features.
Current Limitations Worth Mentioning:
Though Seeing AI excels in various aspects, it does have some limitations:
- The app supports a limited number of languages for OCR recognition compared to competitors like Envision. Additionally, the currency reader feature has a limited range of supported currencies.
- In the light detection feature, the app uses audio tones to convey the amount of light in your environment. Incorporating text information alongside audio tones would be beneficial for users uncomfortable with relying solely on auditory cues.
- The face recognition feature tends to overestimate ages. During testing with three individuals, the app consistently estimated their ages to be at least 3 years older than their actual age.
Final Thoughts:
While other Android apps offer similar features to Seeing AI for visually impaired users, Seeing AI entered the competition with strength and determination. With continued enhancements, this app holds significant potential to evolve into a reliable blind assistant across various aspects. Its current lead in AI image recognition capabilities sets it apart from competitors, at least until Google decides one day to expand its AI Lookout image descriptions globally, or when Be My Eyes completes its availability of the Be My AI feature to the public and adds the capability to share images to the app for recognition.
Seeing AI pleasantly surprised us by arriving as an almost complete, feature-rich package, providing both blind Android users and competitors with reasons to care for what it has and follow its coming steps.
Comments