Skip to content
SIH Buddyby Ganeev Singh
Dev
All problem statements
SIH26173Worth consideringacceptance 3/5

iTantra -Indian Multilingual TTS & STT Aided Neural Transceiver Radio Access for low bitrate links

Indian Space Research Organisation(ISRO) · Miscellaneous · Software

The voice-over-text transceiver is a clever idea with a compelling demo, but ten languages is a heavy multilingual lift where lower-resource languages have weak on-device models — build the loop solidly in two or three languages and be honest about coverage rather than claiming uniform quality across all ten.

What it actually is

Voice is data-heavy and hard to send over low-bandwidth links, but in distress situations voice is more inclusive than text, reaching people regardless of literacy. The ask is an Android app that runs speech-to-text and text-to-speech locally for ten Indian languages, so one phone converts speech to text, sends the tiny text over a low-rate link, and the receiving phone speaks it back — a walkie-talkie that transmits text but behaves like voice.

What to build

An Android app bundling lightweight on-device STT and TTS for ten named Indian languages, where the STT module detects speech pauses to form sentences, transmits only the resulting text over a WiFi or Bluetooth link to another phone running the same app, and the receiving TTS module converts that text back to intelligible speech played as a voice note, with alert messages announced at maximum volume non-interruptibly, working as a push-to-talk walkie-talkie over the low-bitrate link and as a normal channel when push-to-talk is off, evaluated on model footprint, word error rate, TTS legibility and end-to-end latency.

Smallest thing that wins the room

Speak a sentence in one Indian language into one phone, show it transcribed on-device and sent as text over Bluetooth to a second phone, which speaks it back intelligibly in that language — the full walkie-talkie loop — with the end-to-end latency and word error rate reported.

How crowded this one gets

A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.

Moderate150–340 teams expectedroughly 1 in 125–289 wins it

Quieter than 35% of the 226 · #148 of 226 by expected field

A normal-sized field. Your idea has to be good, not miraculous.

Why: defence, intelligence and space bodies drew small fields.

This is a guess, not a fact

Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.

The scores

The number is the shorthand. The line under it is the reason.

What you will be writing

  • On-device STT (Vosk / wav2vec2 / Whisper-tiny)
  • On-device TTS (Coqui / VITS-lite) for Indian languages
  • Pause-based sentence segmentation
  • WiFi Direct / Bluetooth text transport
  • Push-to-talk walkie-talkie logic
  • WER / RTF latency evaluation
  • Speech technology
  • Multilingual on-device AI
  • Low-bandwidth communication

Prior art to read before you start

on-device multilingual STT and TTS · voice-over-text low-bitrate transceiver · push-to-talk speech relay

Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.