Osegi
AI·2 min read

Sarvam AI Gives Lifetime Access to Speech-Impaired Survivor

Sarvam AI CEO Pratyush Kumar has granted lifetime access to its models to a 70-year-old cancer survivor who asked for affordable text-to-speech tools.

A person typing a message on a phone while a text-to-speech app reads it aloud, illustrating AI voice tools for people with speech disabilities.
Image: mint
SUMODialoediviewergiftdetect

Sarvam AI CEO Pratyush Kumar has given lifetime access to the company's AI models to Raghunath AS, a 70-year-old oral cancer survivor who lost his ability to speak. The move followed a public appeal for affordable text-to-speech (TTS) technology for people with speech disabilities.

In a post on X, Raghunath said he has communicated through handwritten notes since his cancer battle in 2015, and that many of those notes go unread or unanswered. "A doctor may not read my 100-word note, but could listen to it in under a minute," he wrote. He cited an estimated 1.99 million people with speech disabilities in India and argued that voice generation is an essential tool for independence, not a convenience.

The pricing complaint

Raghunath pointed to Sarvam's per-character pricing, which he said makes regular use difficult for people with disabilities. He asked the company to consider an affordable, unlimited monthly subscription for Divyangjan, the term used in India for people with disabilities.

Kumar replied on X: "We agree that you shouldn't have to count characters to be empowered by AI. We have credited your account with lifetime access to our models." He also invited other people who use Sarvam's technology for accessibility purposes to contact the company, saying the team would work with them on their requirements. According to Kumar, Sarvam intends to launch a formal accessibility programme in the future.

What Sarvam's TTS offers

Text-to-speech converts written text into spoken audio, the reverse of speech-to-text. Sarvam provides it through its Bulbul V3 model, which the company says produces natural-sounding speech with emotion. Its website lists support for 11 languages: Hindi, Bengali, Tamil, Telugu, Gujarati, Kannada, Malayalam, Marathi, Punjabi, Odia and Indian-accented English, with multiple speaker voices per language.

The gesture covers one user, but it puts a spotlight on how usage-based AI pricing can exclude people who depend on the technology daily. The planned accessibility programme has no announced details yet.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Tags:sarvam aitext-to-speechaccessibilitybulbul v3indiapratyush kumar

More from this day

See all →

More in: AI

See all →