Multiple Wake Words

The Snips platform offers the possibility to run multiple Wake Words in parallel.

If you want to run several universal Wake Words, you will need to edit the snips.toml configuration file. In the [snips-hotword] section, uncomment the model = ... line, and fill it in the following form, with as many models as you want to run:

model = ["<new_path_to_your_model_1>", "<new_path_to_your_model_2>"]

Each path needs to point to a custom_hotword folder, as obtained when downloading an assistant from the Snips Console. sensitivity of both wake words will be set to 0.5 by default. If you want to specify it for each model, use the following syntax:

model = ["<new_path_to_your_model_1>=<sensitivity_1>", "<new_path_to_your_model_2>=<sensitivity_2>"]

If you want to mix universal and personal wake words, please check the personal Wake Word guide.

Impact on footprint and performance

The procedure above will run as many Wake Word engines as indicated. This means that the impact on the footprint will be proportional to the number of Wake Words you aim to detect. Mechanically, the number of false positives will also increase linearly with the number of Wake Words being detected.