🤓 Latest Submissions
Continuum: Coherent Spatial Audio Pipeline
Continuum solves one of the hardest problems in automated spatial audio: temporal coherence across visual cuts. Traditional auto-panning algorithms react to each shot in isolation, causing audio sources to jarringly "jump" between speakers every time the camera angle changes. Continuum fixes this via a novel cross-scene memory engine powered by a multi-agent AI architecture. Our end-to-end pipeline operates in six autonomous stages. First, PySceneDetect segments the video into discrete scenes, while Demucs extracts isolated stems (vocals, drums, bass, other) from the flat stereo mix. Next, a Vision-Language Model (Kimi K2.6 via Fireworks AI) captions keyframes from each scene, providing a detailed description of the visual environment. A reasoning LLM (DeepSeek] via Fireworks AI then acts as a spatial director, assigning each stem to a channel — center, left, surround, overhead, or a full ambient bed — based on those visual cues. Crucially, before finalizing placements, our Coherence Memory Engine cross-references each assignment against a persistent placement history, locking recurring sonic elements to their established channel across cuts unless the scene visibly justifies a change. Finally, placements are compiled into an ADM metadata manifest and rendered into stereo binaural, 5.1, 5.1.4, or 7.1.4 immersive formats using the open-source EBU ADM Renderer. Continuum was originally architected for self-hosted inference on an AMD MI300X via AMD Developer Cloud; we pivoted to Fireworks-hosted inference to keep the pipeline reliable for judged evaluation. Continuum features a Next.js dashboard where creators can monitor their renders and view programmatic visualizations — like coherence timelines and polar position plots — that verify the memory engine is holding placements consistently across cuts. By combining commodity models with our coherence logic, Continuum turns flat stereo into spatially coherent audio without manual labor.
13 Jul 2026
OVATION: Multi-Agent Review Response System
Responding to customer reviews quickly, consistently, and empathetically across Google, Yelp, and TripAdvisor is a constant operational burden for hospitality and service businesses. Generic templates feel impersonal, manual responses don't scale, and a single tone-deaf reply can trigger real reputational damage — yet most automation tools treat every review the same way, with no quality control or escalation path for high-stakes cases. OVATION solves this with a coordinated team of six specialized AI agents — Monitor, Triage, Research, Drafter, QA, and Escalation — built on Band's room-based multi-agent messaging. Each agent runs as an independent Python process with a clearly scoped job: Monitor watches for incoming reviews, Triage classifies sentiment and urgency, Research gathers relevant context (business policies, prior interactions, platform norms), Drafter writes a tailored response, QA evaluates it against nine concrete checks (character limits, empathy, legal safety, brand voice, and more), and Escalation routes anything risky to a human. Every handoff travels through Band's @mention-based routing as structured JSON envelopes, so agents share state and context instead of operating in isolation. To keep the system affordable to run, LLM usage is tiered: premium DeepSeek-V4 (via AI/ML API) powers the higher-stakes Drafting and QA steps, while cheaper Featherless-hosted models handle Triage, Research, and Escalation, and Monitor requires no LLM at all. A live dashboard visualizes the entire pipeline in real time — reviews moving through a kanban-style board, QA scores and check-by-check breakdowns in a detail drawer, and a scenario selector that lets judges trigger five distinct review types on demand, including a full QA revision loop and an end-to-end human escalation path.
19 Jun 2026
