Smallest.ai Review, Pricing & Alternatives (2026)

Smallest.ai is a voice AI company founded in late 2024. It develops Lightning text to speech, Pulse speech to text, Hydra speech to speech, Electron LLM, and a hosted voice-agent platform. TechCrunch reports that the company raised $13 million in Series A funding in July 2026, bringing total funding above $21 million. The round was led by Seligman Ventures, with Sierra Ventures and 3one4 Capital participating. The company announced its Series A on October 5, 2026. Smallest.ai positions the platform for real-time enterprise conversations and customer support. Its product scope makes it a developer and business infrastructure vendor rather than a consumer audio editor or general-purpose chat application. This funding supports work on real-time speech-to-speech technology and enterprise voice conversations.

Smallest.ai exposes its capabilities as separate APIs and hosted agents. A typical API workflow sends text to Lightning for streamed speech or audio to Pulse for transcription. The model overview also lists Hydra for speech to speech and Electron as an LLM. Hosted agents combine these components with web widgets, knowledge bases, testing, post-call analytics, phone numbers, and call pricing. Pulse supports streaming and non-streaming audio, while its API accepts WAV, MP3, FLAC, Opus, μ-law, A-law, and raw PCM. Pulse also offers PII and PCI redaction through request parameters, with original audio not retained after processing. These options let teams choose individual services or assemble a workflow around hosted agent operations.

Lightning v3.1 is the current TTS model described in the documentation. It produces 44,100 Hz audio, accepts 20 language codes plus automatic routing, and uses a proprietary license. The model documentation describes voice cloning and streaming, but does not state a current voice count or cloning duration. Pulse is the current multilingual STT model, with realtime pricing of $0.004 per minute and batch pricing of $0.003 per minute on its Standard Plan. It supports 100 concurrent streaming requests and 25 batch requests per minute on that plan. Pulse 2.0 is described as English-only streaming with turn detection, emotion, and gender. Lightning v2 and lightning-large are deprecated. Native scripts improve Lightning results, while Pulse redaction supports English and Hindi.

Model APIs are metered. Lightning v3.1 costs $0.175 per 10,000 characters on the model pricing page. Pulse costs $0.003 per minute for batch transcription and $0.004 per minute for realtime transcription on the Standard Plan. Voice Agents Pay As You Go starts with $10 in free credits, provides unlimited agents, and charges $0.09 to $0.21 per call minute depending on selected models. Smallest hosting adds $0.01 per minute. The plan has no monthly commitment. Enterprise pricing is custom, with calls listed as low as $0.05 per minute. A phone number costs $10 per month on Pay As You Go, while knowledge-base storage costs $3 per gigabyte per user and queries cost $2 per 1,000. These separate charges require usage estimates before budgeting.

The platform’s main limitation is that its documentation centers on APIs and hosted voice infrastructure rather than a creator-focused editing application. Pulse redaction supports English and Hindi, and Lightning v3.1 requires native scripts for best results. Its proprietary license also rules out self-hosting the model. ElevenLabs is a closer fit for creator-oriented voice production, Cartesia is another developer TTS option, and Deepgram is a direct STT alternative. Buyers should compare voice generation, transcription, agent hosting, deployment, and data-retention requirements separately. Smallest.ai is the best fit for teams that want metered speech APIs and hosted voice agents in one platform, especially when a single vendor is preferred and usage-based billing suits expected demand.