@23017160116943726

TruthTone is an evidence-grounded video captioning studio built for the AMD Developer Hackathon Track 2 challenge. Instead of asking one model to understand a video and produce a joke in a single prompt, TruthTone separates facts from tone. The application first samples important moments from the full video timeline, combining evenly spaced frames with scene-change detection. These timestamped frames are analyzed through Fireworks AI to create a Temporal Truth Spine: the visible subjects, ordered events, visible text, evidence frame references, uncertainty, and the central event of the clip. Next, TruthTone generates three caption candidates for each required style: formal, sarcastic, humorous-tech, and humorous-non-tech. A second AI verification stage compares all candidates against the Truth Spine and scores factual consistency, temporal order, tone match, main-event coverage, and clarity. Captions that introduce unsupported people, dialogue, causes, objects, or outcomes are rejected or repaired before being shown. The result is not simply four prompts. It is one shared evidence system with four distinct voices. The interface makes the process visible through an event timeline, evidence references, quality scores, candidate rejections, repair notes, latency, and token usage. The original video remains in the browser; only selected timestamped frames are sent for analysis. TruthTone proves that tone can change, but truth cannot.

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.