Project: Rangpuri Speech Dataset Collection

Project: Rangpuri Speech Dataset Collection Objective: Create a high-quality speech dataset in the Rangpuri language for AI, Automatic Speech Recognition (ASR), and Text-to-Speech (TTS) model training. Deliverables: * Total Recordings: 10,000 audio clips * Estimated Total Duration: Approximately 16–18 hours * Language: Rangpuri (Bangladesh) * Native Speaker: Male, Age 21 * Recording Environment: Quiet studio-quality environment * Audio Format: WAV (16-bit PCM) * Sampling Rate: 48 kHz * Channels: Mono Quality Specifications: * Low background noise * High Signal-to-Noise Ratio (SNR) * Consistent pronunciation and speaking pace * Natural native accent * Proper segmentation (one sentence per file) * All files manually reviewed for quality before delivery Metadata Included: * Speaker ID * Speaker Age * Speaker Gender * Speaker Region * Clip ID * Recording Duration * Transcript Mapping (if required) Dataset Usage: Suitable for: * Text-to-Speech (TTS) * Automatic Speech Recognition (ASR) * Large Language Models (LLMs) * Voice AI Research * Speech Recognition Systems Final Deliverables: * 10,000 WAV audio files * Metadata (.CSV/.JSON) * Transcripts (if required) * Organized folder structure * Ready-to-use AI training dataset