Skip to content

Downloads stt

The following STT models are available for download. These are compatible with TrulyNatural STT SDK 7.5.0 and later.

Contact your account representative or Sensory sales for additional languages and customizations.

Filename key

opt-vg-vad-stt-
These are pipelines made from the tpl-opt-spot-vad-lvcsr template with a US English "Voice Genie" wake word in slot 0 and an STT recognizer in slot 1.
-B
Model includes an NLU component that identifies intents and entities.
-pnc
Model includes punctuation and capitalization.
-slm
Model includes a small generative language model.

Larger models are more accurate but also require more CPU cycles.

Language Domain File size in MiB Model
English automotive 226 opt-vg-vad-stt-enUS-automotive-large-1.3.14-B-pnc
English automotive 91 opt-vg-vad-stt-enUS-automotive-medium-2.3.14-B-pnc
English automotive 49 opt-vg-vad-stt-enUS-automotive-small-2.3.14-B-pnc
English general 11 opt-vg-vad-stt-enUS-general-micro-2.0.3
English general 7 opt-vg-vad-stt-enUS-general-nano-2.0.3
English general 199 opt-vg-vad-stt-enUS-general-large-2.0.3-pnc
English general 67 opt-vg-vad-stt-enUS-general-medium-2.4.3-pnc
English general 28 opt-vg-vad-stt-enUS-general-small-2.2.3-pnc
German general 199 opt-vg-vad-stt-deDE-general-large-2.2.3
German general 64 opt-vg-vad-stt-deDE-general-medium-2.3.3
German general 25 opt-vg-vad-stt-deDE-general-small-2.3.3
French general 202 opt-vg-vad-stt-frFR-general-large-2.0.3
French general 64 opt-vg-vad-stt-frFR-general-medium-2.3.3
French general 25 opt-vg-vad-stt-frFR-general-small-2.3.3
Italian general 197 opt-vg-vad-stt-itIT-general-large-1.2.3
Italian general 64 opt-vg-vad-stt-itIT-general-medium-2.3.3
Italian general 25 opt-vg-vad-stt-itIT-general-small-2.3.3
Japanese general 215 opt-vg-vad-stt-jaJP-general-large-2.2.3
Japanese general 64 opt-vg-vad-stt-jaJP-general-medium-2.2.3
Japanese general 25 opt-vg-vad-stt-jaJP-general-small-2.3.3
Korean general 215 opt-vg-vad-stt-koKR-general-large-2.2.3
Korean general 64 opt-vg-vad-stt-koKR-general-medium-2.3.3
Korean general 25 opt-vg-vad-stt-koKR-general-small-2.3.3
Spanish general 197 opt-vg-vad-stt-esES-general-large-2.2.3
Spanish general 64 opt-vg-vad-stt-esES-general-medium-2.4.3
Spanish general 25 opt-vg-vad-stt-esES-general-small-2.3.3

Provenance

The wake word, and the speech-to-text acoustic, language, and NLU models are owned by Sensory and have no third-party dependencies.