Last year I was in a night market trying to haggle over a jacket, and Google Translate kept lagging half a sentence behind — by the time the vendor's response was translated, we'd both moved on. It works fine for reading a sign. It falls apart in a live, back-and-forth conversation.

The gap is latency and context. Most translation apps process a full sentence, send it to a generic model, and wait. That round-trip kills the flow of a real exchange — bargaining, asking for directions, joking with a local. Tourists don't need translation, they need a conversation partner.

So I've been training language models specifically for live speech-to-speech translation — tuned on real conversational audio (accents, slang, interruptions), not just clean text. The goal is sub-second turnaround so it feels like a conversation, not a lookup tool.

Built it into TouristLingo. If you travel solo and rely on translation apps, curious what's broken for you the most — lag, accuracy, or just not having it hands-free. Happy to compare notes.