Fluxions

Fluxions: Teaching machines to speak human

Fluxions builds the entire voice stack from the weights up, encompassing speech recognition, reasoning, and speech generation. The models are designed to be small enough to run locally on a phone and cost-effective for business applications.

Key Features

• Complete Voice Stack: Integrates proprietary speech-to-text (akro), reasoning (LLM), and text-to-speech (vui) into a single pipeline.
• Phone Agents: AI agents capable of answering business lines, taking bookings, triaging call-outs, and filling appointment books.
• vui for iPhone: An on-device voice assistant that runs entirely on the iPhone GPU (iOS 26+), ensuring privacy and offline functionality.
• Expressive Text-to-Speech: Supports performed non-speech events like sighs, laughs, gasps, and hesitations.
• On-Premise Licensing: Models can be air-gapped and run on proprietary infrastructure or devices with an annual license.

Pricing

• Voice agents (all-in): $0.10 per minute
• Transcription: $0.20 per hour
• Text-to-speech: $9 per 1 million characters
• vui for iPhone: Free during beta

Use Cases

• Automating front desk operations for dental offices and restaurants.
• Managing call-out desks for plumbing and other service businesses.
• Deploying private, offline voice assistants on mobile devices.

Getting Started

Website: https://fluxions.ai

Fluxions provides a unified API and on-device models for developers and businesses to deploy expressive, real-time voice AI without relying on multiple third-party vendors.