Food recommendation systems share a common dependency: they require an interaction history before they can offer anything useful. First-time users receive no useful recommendations, while recommendations for returning users rely on the previous week’s orders as an approximation of current preferences. Both cases lack important contextual information: weather, mood, and time of ordering are rarely considered, while age and hunger are almost never incorporated. This paper introduces Smart Dine Pro, a system designed to avoid dependence on interaction history. At each request, the system captures the user’s facial expression using DeepFace, retrieves live weather data from the OpenWeather API, and combines these with the user’s age, hunger level, and stated food preference to score menu items using a weighted algorithm. Users who prefer not to use the webcam can instead submit their mood through a short questionnaire, and the same scoring engine processes both input routes. We evaluated the system across 140 input combinations, spanning seven emotional states, five weather conditions, and four age groups. Contextually aligned suggestions were produced in 95.7% of cases, with zero age-restriction violations. A user study involving 30 participants reported an overall satisfaction score of 3.9 out of 5 and a recommendation acceptance rate of 73.3%.