Skip to content

feat: Subtitles for user uploads

Satyapranavanadh Rayapureddy requested to merge implement_subtitles into main

📝 Description: This Merge Request implements Real-time Subtitle Generation and Streaming for audio/video uploads. By bridging the Swecha Gonthuka ASR model with a custom Server-Sent Events (SSE) pipeline, users can now see human-readable Telugu transcriptions appear instantly as pieces of audio are processed.

🌟 Key Features: Real-time Subtitle Dock: A new UI component that displays synchronized Telugu subtitles during transcription. Live SSE Streaming: Implemented a transition-ready /api/subtitle-stream endpoint that processes audio in 30-second chunks and "pushes" text to the frontend immediately. Multi-Format Export: Users can now download transcriptions in standard SRT or WebVTT formats for use in external video players. AI Refinement Pipeline: Integrates Indic Punctuate for natural reading flow. Uses NLLB-200 as a verification layer to ensure consistent Telugu script output.

Merge request reports

Loading