Voila is an open-source family of end-to-end voice-language foundation models designed for real-time, persona-aware conversations. It delivers ultra-low latency full-duplex interactions (~195 ms) and provides a unified model for multiple audio tasks, including automatic speech recognition (ASR), text-to-speech (TTS), and multilingual speech translation with small adaptation. The architecture is a hierarchical multi-scale Transformer that fuses large language model reasoning with acoustic modeling. Voice persona and control can be specified by text instructions; voices can be customized rapidly from short audio samples (as little as ~10 seconds). A Voila Voice Library supports scalable personalization with a large repository of pre-built voices. The project ships with multiple model variants and tooling (Voila-base, Voila-chat, Voila-audio-alpha, Voila-autonomous-preview, Voila-Tokenizer, Voila-Benchmark, Voila-million-voice) and is released as open-source, with code and model weights available on GitHub and Hugging Face.