Speech Recognition Settings

Acoustica relies on Whisper for speech recognition. There are several Whisper models available, both multilingual and for specific languages. The Speech Recognition Settings let you download and manage models. Speech recognition is computationally heavy, and even modern CPUs struggle with the larger models. GPUs are generally better suited, and you can choose which computing device to use for speech recognition. Depending on your system, you can perform speech recognition on either your CPU or GPU.

The Speech Recognition Settings let you download and manage Whisper model files and choose where to perform speech recognition.

The currently selected model is indicated with a bullet and you can click other downloaded models to select another model. The Acoustica installer contains the tiny models for both English and multilingual use. You can click the download buttons to download and install other models. This will open the Acon Digital Installation Manager, which will immediately start downloading the selected models. The model will be available for selection in the Speech Recognition Settings once the download is complete.

If you work with multiple languages and have language specific models installed, you can let Acoustica automatically choose the language specific models since these generally perform better with the same model size. Please click the Automatically select language-specific model if available toggle button to activate or deactivate the automatic selection of language-specific models.