Choosing what to watch can sometimes take longer than watching the movie itself, especially when several people have different tastes. Most streaming platforms make users scroll through endless rows using a TV remote, which can quickly become frustrating. We built Compass, a voice-first TV companion that helps people find something to watch through a natural conversation. You can explain what you are in the mood for, ask for changes like shorter movies or a specific theme, and see the recommendations update as you speak. Compass also has a group mode for finding something everyone will enjoy, and a “Choose for me” mode when you do not want to decide. Each household profile has its own preferences, which are combined with movie data to create more relevant results. SLNG handles speech recognition and voice generation, Unmute powers the real-time voice agent, and Nebius Token Factory handles the conversation and decides when to call our recommendation tools. The backend finds the movies and sends them to the TV interface, so the conversation and recommendations work together in real time.
Compass Recomender System
Compass is a voice-first TV companion that turns natural conversation and household preferences into personalized movie recommendations.
About the project
More from this event
HackBarna AI Summit 26

AI Meeting Moderator
Automatically moderates meeting by voice and keeps track of agenda
By Vitaly Alt, Artem Shambalev

ARCA - wildfire evacuation assistant
ARCA shows emergency teams who and what a wildfire threatens, how long evacuation will take, and where to act first
By Shahnawaz Akhtar, Balazs Lapis, Melissa Orphali, Viktoria Sargsyan

ARLANDRIA
Eliminate hallucinations expert-domain areas (e.g. science) by limiting LLMs to language processing instead of tool usage
By Caetano Souto Maior