📋 Design Prompt
Add voice after the text workflow works
Updated October 31, 2026. Educational guidance; use current official documentation and your installed version.
For built-in dictation, prompt cleanup, and voices as context cues, start with Voice in, voices out. Dictation does not require this custom-output setup.
For a concrete marker-to-queue-to-speech implementation, see Chatterbox and completion hooks. Inspect and test it before installing hooks.
Voice output is optional. It introduces model downloads, audio hardware, latency, provider terms, and another service to supervise. A voice demo does not prove the whole assistant is operational.
Help me add speech output to one verified text workflow.
Inspect available supported TTS tools and current official documentation.
Offer a local or hosted option based on my hardware, privacy needs, and budget.
Use a voice I own or have permission to use; do not assume cloning rights.
If I have several roles, propose a stable role-to-voice mapping and explicit role labels.
Use short spoken summaries, queue competing updates, and retain written results.
Confirm the playback destination, quiet hours, and what may be read aloud.
Do not change global audio, payment, or provider settings without authorization.
Generate a short test, inspect its duration/content, and play it on the intended device.
Verify audible playback, not only file creation or HTTP200.
Keep text fallback, bounded timeouts, and no silent paid fallback.
Document restart recovery and remaining limits.