Skip to main content

Speech recognition models

Choose the model to use. We provide the default model sommers and whisper, a Korean fine-tuned Whisper model. When using whisper, you must set the language parameter.

Whisper supports 100 languages. If you want language detection, see Language.

caution

Default concurrency is 2 and can be adjusted upon request. For Whisper-related inquiries, contact us via chat below.

Sommers model​

  • Our in-house ASR model with fast latency and strong accuracy.
  • Supported languages: Korean (ko) and Japanese (ja).
  • language is optional. If omitted, it defaults to ko (Korean).

Korean Fine-tuned Whisper Model​

  • Fine-tuned for stronger Korean speech recognition while retaining Whisper's multilingual and language detection capabilities.
  • language is required. For detection/multi-lingual usage, see Language.

Usage​

In the POST Request body, set model_name to sommers or whisper (whisper requires language).

POST example: Example

Unified example (presets): Example