elbrad0/night-road-narrator

Night Road Narrator is a local, offline AI voice narrator built for a specific Steam game, Vampire: The Masquerade, Night Road. It reads the game's on-screen story text aloud as you play, giving the narrator one consistent voice and each character their own distinct voice. Everything runs on your own computer, with no cloud services, no API keys, and no account needed. The project is not affiliated with the game's makers and does not touch or modify any game files, it only reads the visible text and plays audio alongside it. The tool figures out which lines are dialogue and which are narration, matches character voices to their apparent gender based on names and pronouns in the text, and keeps each character's voice consistent for the whole session. It never cuts off long passages, correctly skips over stat and menu screens while still reading any story text on those pages, and lets you skip ahead by simply clicking to the next part of the game, which stops the current narration and jumps to the new text. Volume can be adjusted with keyboard shortcuts during play, and the chosen level is remembered for next time. Numbers, times, and dates are read out in natural spoken form rather than digit by digit. Under the hood, four pieces work together: the game runs as normal with a special launch option enabled, a small script watches for new on-screen text, a local AI model run through Ollama decides who is speaking each line, and a program called Piper turns the labeled text into speech. To use it you need Windows 10 or 11, since the one-click launcher is Windows-only, though the underlying Python code could be adapted to other systems. You also need the game itself on Steam, roughly 10 gigabytes of free disk space, Python, Ollama, and Piper. A dedicated graphics card is strongly recommended, since the speaker-detection AI and speech synthesis run locally on your own machine, though it can run on a processor alone with more delay. Setup involves downloading the project, installing Python and a handful of Python packages, installing Ollama and pulling an AI model, downloading a single voice model, building a voice gender lookup file, and turning on a debug option in the game's Steam launch settings. After that, a single batch file starts everything and launches the game for you.