Smallest AI
Ultra-low-latency TTS and STT APIs with top-ranked real-time Hindi and Hinglish voices.
About
Smallest AI offers real-time speech APIs for developers: Lightning text-to-speech and Pulse speech-to-text. Lightning v3.1 streams its first audio chunk in about 80-100ms, supports 12-31 languages depending on the pool (including Hindi with natural code-switching), offers voice cloning, and is available over HTTP, SSE, and WebSocket. Pulse STT reports 64ms time-to-first-token with speaker diarization and emotion detection.
Independent Voice Arena listeners ranked Lightning first among real-time models for Hindi customer support in mid-2026, ahead of ElevenLabs, Cartesia, and Sarvam. The company provides official docs, a Vercel AI SDK provider, and agent skills, with TTS priced around $0.195 per 10k characters on the standard plan.
Limitations: arena rankings reflect blind listener votes on specific categories rather than universal quality; per-character pricing can add up fast for high-volume telephony use.
What it can do
- Lightning TTS with ~80ms streaming first-audio latency
- Pulse STT with diarization and emotion detection
- Voice cloning and 12+ language support
- Vercel AI SDK provider and official agent skills
Setup
Get an API key at app.smallest.ai; POST to api.smallest.ai/waves/v1/tts with the key as a bearer token.
Prerequisites
A Smallest AI API key.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 80; Trending score 76/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.