loading
loading
When Seema Kharad, a visually impaired student in Pune, was preparing to pursue a PhD in political science, finding accessible study material was a challenge. Many of the books and academic documents she needed were unavailable as audiobooks, leaving her dependent on friends to read PDFs and notes aloud. For Yogen Ghodke, an AI solutions engineer based in New York, this presented an opportunity to use technology to address a practical barrier to education. He began building ListenToMe, an app that converts study material into audio, with Kharad testing early versions and identifying problems. Kharad’s father introduced them, and Ghodke began working on a solution, asking her about the specific challenges she faced and what would satisfy her needs. He built an initial version of the app using AI and spent around two months working with Kharad to incorporate her feedback. As the product improved, they began exploring whether it could help others facing similar challenges. The app has since attracted users beyond India. His thinking was also shaped by his experience of accessibility provisions in the US. Ghodke said he was struck by how universities accommodated people with disabilities through features such as wheelchair ramps, accessible parking, and button-operated doors. He began considering how a similar focus on accessibility could apply to AI-powered education, particularly for students who cannot independently access conventional learning material. Building for real-world use ListenToMe uses optical character recognition (OCR) to extract text from documents and text-to-speech technology to convert it into audio. Users can upload PDFs, photos of documents, handwritten notes, and other supported file formats from their phones. The system converts the extracted text into audio and makes it available to users. Ghodke experimented with several open-source AI models but found that they required further fine-tuning. He relies on user feedback to identify errors, particularly in languages he does not understand. “Recently, for example, we had a problem where figure numbers were not being pronounced correctly,” he said. “So instead of saying 100, it was saying one zero zero.” Ghodke said he makes corrections regularly as feedback comes in. According to him, the app supports around 30 languages, including English, Marathi, Hindi, Tamil, Telugu, Assamese and Nepali. It has attracted users from countries including the US, Azerbaijan and Tunisia. To improve discoverability, Ghodke added search keywords and descriptions in different languages to the app’s Google Play Store listing. Users have also shared it in WhatsApp groups for blind people. Some have requested additional language support. After a user from Iran began processing Persian documents, Ghodke added support for the language. The user then shared the app with friends, helping it reach a wider audience. “I can add support for any language,” he said, describing how user requests have shaped the app’s development. ListenToMe can also be used beyond academic settings. People with low vision can photograph a newspaper and have its contents read aloud. A translation feature allows users to listen to English-language material in Marathi or another supported language. Growing usage, managing costs At the time of the interview, around 1,600 people had signed up for ListenToMe, and about 1,000 had submitted documents for processing, according to Ghodke. The app had been publicly available for only three or four weeks, and he said its user base was growing rapidly. “It is very much growing very fast, and it’s almost growing too fast,” he said, describing the pace at which new users were joining. He added that the app could see around 200 new users a day. The app is currently free, with a daily processing limit of 250 pages. Ghodke is considering a plan priced at around ₹149 that would allow users to process up to 500 pages a day. Some users have asked for higher limits and indicated that they would be willing to pay a reasonable price, he said. For now, Ghodke is funding the operating costs himself and runs the service on a Mac Mini. He estimates the current processing cost at around one US cent per page and wants to bring it down to ₹1 for every 100 pages. The limits are intended partly to prevent a small number of users from consuming computing resources needed by others. “If one user is uploading thousands of PDFs in a day, then other users will not get any resource,” he said. Questions around document privacy Because ListenToMe processes user-uploaded documents, data handling is a key consideration, especially for people submitting academic notes or personal material. Ghodke said uploaded PDFs are stored on his server while the audio is generated. The system deletes the original documents once processing is complete, while generated audio files remain available for a few days before being automatically deleted. The temporary storage period lets users download their files while limiting how long they stay on the server, Ghodke said. Making accessible education more widely available Ghodke’s longer-term plans include formalising the business, raising funds and hiring people to improve the models. He also wants to reach more universities, student organisations and NGOs that could help introduce the app to people who need it. He is particularly interested in reaching schools in India, where affordability could determine how widely the technology is adopted. One model under consideration would involve NGOs helping fund access for people who cannot afford paid services, while subscriptions in markets such as the US help cover operating costs. Faster GPUs and more optimised models could increase processing capacity and allow the service to support more users, he said. For now, securing funding for infrastructure and technical development remains a priority. “The building part is done. Now, the next part is all of that business side,” he said. Ghodke is exploring funding from venture capital firms and organisations that support people with disabilities. He is also working out how to establish a startup in the US while still a student.
document fetched 2026-10-24 11:25