SPEECHVERSE — An ACM GRIET AI Hackathon

Core Focus Areas (Speech Processing): Speech-to-Text (ASR): Low-resource Indian languages Accent-robust ASR Noisy environment transcription Text-to-Speech (TTS): Emotional speech synthesis Regional language voice cloning Assistive voice generation Speech Emotion Recognition: Mental health indicators Call-center analytics Classroom engagement analysis Speaker Recognition: Speaker identification / verification Fraud & spoof detection (deepfake audio) Audio Intelligence: Keyword spotting Voice command systems Audio event detection ️3-Day Hackathon Schedule: Day 1 – Ideation & Foundations: Morning: Inauguration + Keynote“AI & the Future of Speech Technologies” Problem statement release Afternoon: Team ideation & mentor allocation Dataset briefing (LibriSpeech, Common Voice, Indic datasets) Tool onboarding (Whisper, wav2vec2, SpeechBrain, OpenAI APIs optional) Evening: Idea validation checkpoint Shortlisting within teams (mentor feedback) Day 2 – Build & Optimize: Morning: Model building (ASR / TTS / SER etc.) Feature extraction (MFCC, spectrograms, embeddings) Afternoon: Optimization & fine-tuning Real-world constraint handling (noise, accents) Evening: Mid-review (20% weightage) Debug + improve Day 3 – Demo & Evaluation: Morning: Final integration Demo prep (live or recorded) Afternoon: Final presentations (5 + 3 mins Q&A per team) Evening: Results, prizes & networking