halleyfy-audiolisted
Install: claude install-skill supasanahq/halleyfy-skills
# halleyfy audio
Three commands work with a voice:
- `halleyfy speak` turns text into an .mp3, in one of 21 voices.
- `halleyfy revoice` says a recording's words again in another voice, with the same timing. From a video it also saves the video with the new voice.
- `halleyfy isolate` pulls a clean voice out of noisy sound or a video.
A fourth, `halleyfy sound`, makes sound effects and soundtracks (see "Sound effects"). A fifth, `halleyfy music`, makes music (see "Music").
halleyfy picks the model. You write the line and choose the voice. The voice reads the words you give it, so the text is the whole brief: write it the way it is said, with the pauses and the delivery you want.
## Before the first run
1. `command -v halleyfy`. Missing: if `npx` exists, use `npx -y halleyfy` wherever this skill says `halleyfy`. Otherwise tell the user to install it with `curl -fsSL https://halleyfy.com/install.sh | sh` and stop.
2. `halleyfy whoami --json`. Not logged in: run `halleyfy login --no-wait --json`, give the user its `verification_url` and `user_code` (the same link signs a new user up), and after they approve run `halleyfy login --wait --json`. Never type a password or ask for a key.
3. Read `halleyfy speak --help` once: every flag, the voices and the live price. Read `revoice --help`, `isolate --help`, `sound --help` or `music --help` before the first run of any of them.
Don't pass `--local` or `--live` unless the user asks.
`revoice` and `isolate` take sound from a video