Jival Raghav@Jival

Caption Craft AI operates as a multi-stage, closed-loop visual processing pipeline designed to translate raw video keyframes into highly stylized, factually accurate descriptions. The process begins with segment-based keyframe extraction and perceptual image hashing to deduplicate visual frames, which are then passed to a Vision-Language Model (Kimi-k2p6) to generate a structured "Scene Brief" containing the setting, subjects, primary actions, timeline events, and a calibrated "jokeability" rating (low, medium, or high). In the second stage, the agent concurrently spins out four distinct caption styles—Formal, Sarcastic, Humorous Tech, and Humorous Non-Tech—which are then fed into a strict Devil's Advocate QA pass. This critique step acts as a final filter that cross-checks the draft captions against the visual grounding facts, scoring them on factuality and tone match; it actively flags and rewrites first-person violations (banning pronouns like I, me, my, we, and you) and fleeting-subject errors (where a joke is built on a minor background object rather than dominant visual subjects), utilizing a fail-safe multi-provider routing (Qwen3p7 or Gemma 26B) to guarantee that only clean, corrected, and compliant captions are written to the final output file.

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.