Providers

NVIDIA

STT TTS S2S 15 models 101 voices based in United States self-serve API data from documentation synced 2026-11-18

NVIDIA (Santa Clara, California; founded 1993) publishes speech models under its Riva and Nemotron Speech lines: Parakeet, Canary and Nemotron ASR/Transcribe for speech recognition, Magpie for TTS and Nemotron VoiceChat for full-duplex speech-to-speech. Hosted trial endpoints are available with a free API key on build.nvidia.com; production deployment is via self-hosted Speech NIM containers governed by NVIDIA's software licence (NVIDIA AI Enterprise). Many of the models are also released as open weights on Hugging Face.

Speech-to-text 12

CompareModelLangsFeaturesPriceStatus
nvidia/canary-1b-asrCanary 1B Multilingual Many 1/11 –
nvidia/nemotron-transcribeNemotron 3.5 Transcribe 40 3/11 –
nvidia/nemotron-asr-streamingNemotron ASR Streaming 36 4/11 –
nvidia/parakeet-1_1b-rnnt-multilingual-asrParakeet 1.1B RNNT Multilingual 23 3/11 –
nvidia/parakeet-ctc-0_6b-asrParakeet CTC 0.6B English 1 2/11 –
nvidia/parakeet-ctc-0_6b-zh-cnParakeet CTC 0.6B Mandarin-English 2 2/11 –
nvidia/parakeet-ctc-0_6b-esParakeet CTC 0.6B Spanish-English 2 3/11 –
nvidia/parakeet-ctc-0_6b-zh-twParakeet CTC 0.6B Taiwanese Mandarin-English 2 1/11 –
nvidia/parakeet-ctc-0_6b-viParakeet CTC 0.6B Vietnamese-English 2 3/11 –
nvidia/parakeet-ctc-1_1b-asrParakeet CTC 1.1B English 1 3/11 –
nvidia/parakeet-tdt-0_6b-v2Parakeet TDT 0.6B v2 1 2/11 –
nvidia/parakeet-tdt-0_6bParakeet TDT 0.6B v3 25 3/11 –

Text-to-speech 2

CompareModelLangsFeaturesPriceStatus
nvidia/magpie-tts-multilingualMagpie TTS Multilingual 12 1/8 –
nvidia/magpie-tts-zeroshotMagpie TTS Zeroshot 12 1/8 –

Speech-to-speech 1

CompareModelLangsFeaturesPriceStatus
nvidia/nemotron-voicechatNemotron 3 VoiceChatBeta 1 0/19 – Beta

Voices 101

Browse and filter all of them.

Esc

↑↓ move↵ open