Pulse and Pulse Pro transcription.
4 integrated models · Official documentation ↗
Smallest.ai's current standard text-to-speech model, offering 217 voices at 44.1 kHz with trained voices in 12 of its 20 accepted languages, instant voice cloning, and about 200 ms time to first byte when called in-region.
Premium tier of Smallest.ai's Lightning v3.1, served on dedicated capacity with curated Indian, British and American voices, Hindi-English code-switching and 31 languages, at the same latency and request limits as the standard model.
Smallest.ai's multilingual Pulse speech-to-text, its main model for both streaming and file transcription, covering Indic, European and East Asian languages with code-switching, diarization and redaction. Pulse 2.0 now adds built-in turn detection for English streaming.
English-only, file-only tier of Smallest.ai's Pulse line, which the company reports at 5.42% average WER on the Open ASR Leaderboard's ESB sets, tied for second. It does not return speaker labels.