NexGen MediaCraft: Real-Time AI News Anchoring with LLM-Based Generation and Web Scraping
Fuente:
Zenodo
Gespeichert in:
| Hauptverfasser: | , |
|---|---|
| Format: | Recurso digital |
| Sprache: | Englisch |
| Veröffentlicht: |
Zenodo
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866902018699821056 |
|---|---|
| author | HAMZA, AMEER Mairaj, Nimra |
| author_facet | HAMZA, AMEER Mairaj, Nimra |
| contents | <p>This paper presents NexGen MediaCraft, an autonomous AI-powered news anchoring system that combines real-time web scraping, semantic vector search, large language model (LLM)-based text generation, and immersive 3D avatar delivery. The system automatically collects news articles hourly from multiple sources using a Python-based scraper, preprocesses the content, and indexes it in a vector database for efficient semantic retrieval. A fine-tuned DeepSeek LLM generates up-to-date news scripts dynamically, responding to user queries with contextually relevant summaries. A lightweight Flask backend orchestrates scraping, storage, vector queries, and LLM inference, and integrates seamlessly with the Convai conversational API to handle speech synthesis and real-time lip-sync animation. The final news delivery is rendered through a web-based interactive 3D anchor built with React Three Fiber, providing users with a realistic, query-driven news experience. Prototype results demonstrate that the pipeline can deliver consistent, low-latency responses and maintain natural, continuous interaction. NexGen MediaCraft highlights a practical, modular approach to deploying scalable, AI-driven news broadcasting with minimal human intervention.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_15856940 |
| institution | Zenodo |
| language | eng |
| publishDate | 2025 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | NexGen MediaCraft: Real-Time AI News Anchoring with LLM-Based Generation and Web Scraping HAMZA, AMEER Mairaj, Nimra AI News Anchor, Conversational AI, Large Language Models, Web Scraping, Vector Database, DeepSeek, Flask API, React Three Fiber <p>This paper presents NexGen MediaCraft, an autonomous AI-powered news anchoring system that combines real-time web scraping, semantic vector search, large language model (LLM)-based text generation, and immersive 3D avatar delivery. The system automatically collects news articles hourly from multiple sources using a Python-based scraper, preprocesses the content, and indexes it in a vector database for efficient semantic retrieval. A fine-tuned DeepSeek LLM generates up-to-date news scripts dynamically, responding to user queries with contextually relevant summaries. A lightweight Flask backend orchestrates scraping, storage, vector queries, and LLM inference, and integrates seamlessly with the Convai conversational API to handle speech synthesis and real-time lip-sync animation. The final news delivery is rendered through a web-based interactive 3D anchor built with React Three Fiber, providing users with a realistic, query-driven news experience. Prototype results demonstrate that the pipeline can deliver consistent, low-latency responses and maintain natural, continuous interaction. NexGen MediaCraft highlights a practical, modular approach to deploying scalable, AI-driven news broadcasting with minimal human intervention.</p> |
| title | NexGen MediaCraft: Real-Time AI News Anchoring with LLM-Based Generation and Web Scraping |
| topic | AI News Anchor, Conversational AI, Large Language Models, Web Scraping, Vector Database, DeepSeek, Flask API, React Three Fiber |
| url | https://doi.org/10.5281/zenodo.15856940 |