duby_skill

Overview

This skill converts text into natural-sounding speech using the Duby.so API. It supports multiple voice IDs and selectable emotions so you can tailor the output tone. The skill returns a media file path for the generated audio, ready for playback or storage. It is implemented as a simple wrapper that calls the Duby API and saves the result locally.

How this skill works

You provide the text and optionally a voice_id; the skill calls a tts.sh helper that uses curl and jq to interact with Duby.so. The API returns an audio file which the skill saves and returns as a media path. An API key must be provided by setting the environment variable used by the script. Voice IDs map to specific speakers and emotional settings supported by Duby.

When to use it

• Generate spoken narration for videos, podcasts, or tutorials without recording a human voice.
• Create voice prompts or system messages for bots, IVR, or accessibility features.
• Produce quick prototypes of audio UI behavior using different voices and emotions.
• Archive or back up spoken versions of written content for preservation or review.

Best practices

• Provide concise, well-punctuated text to improve prosody and pausing in the generated speech.
• Specify a voice_id when you need a consistent speaker or particular emotional tone.
• Set and protect your API_KEY in a secure environment variable; do not hard-code it in scripts.
• Validate output files and maintain a naming scheme and storage location for generated media.

Example use cases

• Convert blog posts into audio files for a podcast feed using a calm, narrated voice.
• Create UI voice prompts for an assistant with an energetic or empathetic emotion setting.
• Batch-generate spoken versions of documentation for offline distribution or archives.
• Quickly demo different voice styles by swapping voice_id values in test calls.