smallest.ai

Smallest.ai is a web and API platform for enterprises that need low-latency voice AI models and production voice agents for customer conversations in industries like finance, lending, insurance, and healthcare.

Key capabilities include:

• Lightning V2 (text to speech) for multilingual speech synthesis with low latency (time-to-first-byte around 100 ms), 50+ languages, 100+ voices, accent support, and voice cloning with emotional expression controls.
• Pulse (speech to text) for transcription across 38+ languages with streaming and batch modes, code-switching, and detection for emotion, speakers, and timestamps.
• Voice agents and telephony tooling designed for high-volume conversations, including real-time speech and text interactions, auto-dialer support, and analytics, backed by a small language model optimized for conversational use.
• Developer and workflow options via REST APIs, Python/Node.js SDKs, and a no-code workflow builder with integrations (including CRMs, Zapier, and Make.com).

Pricing: Paid, with pay-as-you-go plans for usage-based billing.

What makes it notable: the product focuses on small, efficient multi-modal models (under 10B parameters) built for low GPU usage and very low latency (including ~45 ms time-to-first-token for its Electron SLM), plus support for and enterprise compliance claims including .