
A huge thanks to our customers, partners, investors, and the team. We're just getting started.
We have passed the Turing test for Voice AI and raised $13M in Series A funding to bring the next generation of Voice to enterprises.

Real-time Voice AI. Built for Scale.

A huge thanks to our customers, partners, investors, and the team. We're just getting started.
We have passed the Turing test for Voice AI and raised $13M in Series A funding to bring the next generation of Voice to enterprises.

Which one sounds more natural? One is Lightning v3.1 Pro, and the other is Fish Audio S2.1 Pro. A or B? Which one would you pick?

my therapist told me to draw my feelings i can't draw so i built a voice agent that draws for me 🎨

Hope you're feeling better after drawing.
my therapist told me to draw my feelings i can't draw so i built a voice agent that draws for me 🎨

We are redefining how we hire @smallest_AI We don’t read resumes. There’s nowhere to upload one. Everyone’s shipping AI-written code. Nobody wants the pager. We’re hiring the engineer who does. Software Engineer: you’re the QA, DevOps and SRE team. No resume. No LeetCode. One shot. Bengaluru · on-site · Apply in 2 minutes using this link - hiring.smlst.co


Before you automate a retail or e-commerce call, check that it’s worth it. Where do voice agents actually pay off in retail? Our ebook breaks down the calls worth automating first, the economics behind them, and what it takes to hold up at production volume. Get the free ebook: smallest.ai/ebooks/voice-age…

Guess which one sounds most natural. Three TTS models read the same apology. Which one do you think nailed it? Drop your guess in the replies - A, B or C.

Some voices make it feel like home. It might be a call from family, or a friend you haven’t heard from in a while. This Diwali, we hope you hear a lot of them. Happy Diwali from all of us at @smallest_AI

Lightning v3.1 Pro's new checkpoint is live now. Try it in Hindi by setting the model to lightning_v3.1_pro_10_26: docs.smallest.ai/model-cards…

Our newest Lightning 3.1 Pro checkpoint is the #1 real-time text-to-speech model in Hindi on the @voicearena_ai TTS Leaderboard. It ranks #1 in Hindi, ahead of every other real-time TTS model. In US English, it ranks #3 of 16 among real-time models.
Smallest AI's new Lightning 3.1 Pro checkpoint ranks #3 of 16 in US English 🇺🇸 and #1 of 10 in Hindi 🇮🇳 among real-time models on the Voice Arena TTS Leaderboards. Lightning 3.1 Pro from @smallest_AI is a real-time TTS model built for scale. Every ranking on Voice Arena comes from blind, head-to-head votes by vetted native speakers, scored with Bradley-Terry Elo. The new checkpoint scores 1046 Elo in US English and 1106 in Hindi, up 90 and 87 Elo on the previous Lightning 3.1 Pro, with clear statistical separation from it on both boards. With all models counted, it ranks #7 of 24 in US English and #2 of 21 in Hindi. Results backed by 1,469 blind, head-to-head listener votes from native US English and Hindi speakers. See below for both boards, the exact audio raters judged and unedited listener verdicts 🧵

A practical guide to building voice agents that can handle enterprise-scale call volumes. Download the free ebook: smallest.ai/ebooks/building-…

The hundredth simultaneous call is where most voice agents break. At 1000 concurrent calls, a response that takes 400 ms with one caller can turn into a ~2-second wait. Our new free ebook breaks down the three ways enterprises build voice agents and how to choose the right one for your expected call volume. Get it here: smallest.ai/ebooks/building-…

On the details callers notice, Lightning v3.1 Pro ranks first among 10 models for breathing, natural pace, pause placement and emotion. In our accuracy test, it didn’t skip or invent a single word. Try the preview now. Set the model to lightning_v3.1_pro_10_26: docs.smallest.ai/model-cards…

Full results for all 11 models, by language: docs.smallest.ai/model-cards…

Lightning TTS vs 11 leading TTS models. We put our newest Lightning v3.1 Pro checkpoint head-to-head against 11 TTS models across US English, UK English, Latin American Spanish and Hindi-English. It came out ahead of every model overall. The new checkpoint is in preview now for English, Spanish and Hindi. To try it, set the model to lightning_v3.1_pro_10_26. Full results in the first reply.