Raven-1

Voxtral TTS is a state-of-the-art, multilingual text-to-speech model designed to generate highly realistic and emotionally expressive speech from text, combining strong contextual understanding with advanced speaker modeling to produce natural, human-like audio output. Built as a lightweight model with around 4 billion parameters, it delivers efficient performance while maintaining high quality, enabling scalable deployment for enterprise voice applications. It supports nine major languages and diverse dialects, and can adapt to new voices using only a short reference audio sample, capturing not just tone but also rhythm, pauses, intonation, and emotional nuance. Its zero-shot voice cloning capabilities allow it to replicate a speaker’s style without additional training, and it can even perform cross-lingual voice adaptation, generating speech in one language while preserving the accent of another.

Velma is a voice-native AI model developed by Modulate as part of a broader voice intelligence platform, designed to understand conversations directly from audio rather than relying on text transcripts. Unlike traditional systems that convert speech into text and analyze it with language models, Velma uses an Ensemble Listening Model (ELM), a specialized architecture that processes multiple dimensions of voice simultaneously, including tone, emotion, pacing, intent, and behavioral signals. This allows it to capture the full meaning of a conversation, not just the words spoken, recognizing nuances such as stress, deception, sarcasm, or escalation in real time. It operates by combining hundreds of specialized detectors, each focused on specific aspects of speech like emotional state, inappropriate conduct, or synthetic voice indicators, and then fusing those signals into higher-level insights about what is happening in a conversation.