Siddharth Goutam@siddharthws2694389

This project is an AI-powered video captioning agent built for the AMD Developer Hackathon: ACT II. Given a short video clip, the pipeline extracts evenly-spaced frames, sends them to a vision-language model (Qwen 3.7 Plus, hosted via Fireworks AI) to generate a factual description of the scene, then rewrites that description into four distinct caption styles: formal, sarcastic, humorous-tech, and humorous-non-tech. The core challenge was making each style genuinely distinct rather than just varying the same joke — this was solved through carefully tuned prompt templates per style, combined with disabling the model's default "reasoning" mode (which otherwise leaked internal chain-of-thought into the final captions) using Fireworks' reasoning_effort parameter. The agent is fully containerized with Docker, reads a batch of video tasks from /input/tasks.json, and writes structured results to /output/results.json, matching the hackathon's evaluation harness format. It was tested against varied content — traffic scenes, animals, and office settings — to confirm the captions stay accurate and stylistically distinct across different subject matter.

To participate in the hackathon, simply click Sign up with AMD. If you are not already a member of the AMD AI Developer Program (ADP), you will need to create an account. ADP members can access AI tutorials, experts, and community support. In addition, new members can claim a free month of DeepLearning.ai pro, $100 in AMD GPU cloud credits, and $50 in Fireworks AI API credits. All hackathon participants will receive on hackathon launch additional compute and API credits for their projects. Ready to build what's next on AMD? Sign up with AMD below to secure your spot.