One of Sambhav's early users was a nine-year-old who had gone an entire term without speaking in class unless pushed. He could read perfectly well; he simply would not do it while others were listening. Given a way to practise on his own, his reading improved steadily over the weeks that followed.
He is not unusual. In many classrooms, the children who most need to practise spoken English are the least likely to try. Speaking up means risking a mistake in front of everyone. Vidushee Shekhar built Sambhav to give those children a quieter place to start.
For children who do not speak English at home, that hesitation adds up. Confidence in spoken English often shapes access to further study and work, yet in a crowded classroom there is little time for any one child to read aloud and learn what to fix.
How it works
Sambhav combines structured reading practice with AI-based speech analysis to help children strengthen their phonological skills. A child selects a practice exercise and reads the displayed text aloud, with the app offering more than 100 practice lines that build from individual words to full paragraphs. The system analyses the recording at the phoneme level, examining individual speech sounds rather than relying solely on whether an entire word or sentence has been recognised correctly. A progress dashboard helps track performance across practice sessions, while exercises can be adapted to the child’s learning needs.
Early testing revealed that children could be penalised for how they spoke rather than for the specific reading difficulties Sambhav was designed to address. Vidushee refined the practice material to use more familiar vocabulary and focused the assessment on sound-level patterns, aiming to make feedback more relevant to each child’s learning needs.
From prototype to classroom
Sambhav is now used by more than 12,000 children through The Creative Thinkers Forum (TCTF), where it runs daily at three centres, and by over 3,000 through Nisarg Foundation. In one group, a student's scores rose from 54 to 91 percent.
Vidushee is Sambhav's founder and lead developer, and a patent application for the platform has been published. She built and supported it while studying in Bengaluru, far from the Delhi classrooms where it is used.
Sambhav grew out of Vidushee’s research into how AI evaluates pronunciation. Her research compared classical statistical methods with neural models for phonological error detection, examining their accuracy in evaluating children’s speech, particularly that of second-language English learners. Classical systems typically measure speech against fixed acoustic patterns and can struggle with accents they were not built around, while neural models learn from far larger and more varied sets of recordings.
She presented this work at the international research conferences APSIPA ASC and CHIRA. These findings inspired Sambhav, which works to ensure that how a child speaks does not determine how well they are understood.
Vidushee continues to develop Sambhav alongside her final year of school. For the children who use it, its value is simple: a private space to make mistakes, see what went wrong, and try again until reading aloud no longer feels out of reach.
Sambhav is free to use at https://sambhav-ai.org/
Disclaimer: This is a sponsored article. All possible measures have been taken to ensure accuracy, reliability, timeliness and authenticity of the information; however Outlookindia.com does not take any liability for the same. Using of any information provided in the article is solely at the viewers’ discretion.
