We are building a multi-tone video captioning pipeline for Track 2 of the AMD Hackathon, using Fireworks AI to run fast vision-language models.
Aura-VLC captions any 30s–2min video in four styles—formal, sarcastic, humorous-tech, and humorous-non-tech—using a 3-stage pipeline: Qwen-VL-Max describes, Kimi K2.7 audits frames, and Qwen3.7-Plus stylizes. Built for AMD Hackathon Track 2.