Към съдържанието
ВходЗапочни безплатно

Voice, music and sound effects

Read a script aloud, compose a short track or make a sound effect.

Use it

Copy a skill into your assistant's skills folder. For Claude Code that is ~/.claude/skills/<name>/SKILL.md:

mkdir -p ~/.claude/skills/pickshot-voice-and-music
curl -L https://pickshot.ai/developers/skills/pickshot-voice-and-music/SKILL.md \
  -o ~/.claude/skills/pickshot-voice-and-music/SKILL.md

Download SKILL.md · MCP setup · All skills

The whole SKILL.md

---
name: pickshot-voice-and-music
description: Make speech, music or a sound effect on the user's Pickshot account: read a script aloud in a chosen language, compose a short track, or make a sound effect. Use it when the user wants a voiceover, narration, background music, a jingle or a sound for a clip.
---

# Voice, music and sound effects

This skill needs the Pickshot MCP server (`get_balance`, `list_models`, `estimate_cost`, `create_audio`, `get_job`). If they are missing, tell the user to connect it first (https://pickshot.ai/en/developers/mcp).

The spending rules of the `pickshot-generate` skill apply: quote first, wait for yes, cap the price, one key per request.

## Choose the mode

| The user wants | `mode` | Main input |
| --- | --- | --- |
| A script read aloud | `tts` | `text`, `lang` (`en`, `ko`, `ja` or `zh`), `voice` if they pick one |
| A song or background track | `music` | `prompt` (genre, mood, tempo), `lyrics` if they have them, `instrumental` for no vocals, `seconds` |
| A sound effect | `sfx` | `prompt` (what makes the sound), `seconds` |

## Steps

1. Call `list_models` with `kind` `audio`. Keep models that are `available` and offer the mode. Read the limits: `chars` for the longest script, `seconds` for lengths, `speed` if speech speed can change.
2. For speech, get the exact words. Keep them within the character limit; if the text is longer, offer to split it into parts and quote each part. Ask which language it is in if it is not obvious. Never read out a real person's name in a way that imitates their voice: Pickshot's voices are not clones of people.
3. For music, ask for the genre, the mood and the length. Keep to the lengths the model lists.
4. Call `estimate_cost` with `kind` `audio`. Tell the user the model, what it will make and the credits. Wait for yes.
5. Call `create_audio` with the same request, `max_credits` equal to the quote and a new `idempotency_key`.
6. The answer is a job. Call `get_job` every 10 to 20 seconds until it is `done` or `failed`. Give the user the link in `results[0].url`.

## Notes

- Speech in one language with words from another often sounds wrong. Write the script in the language you set.
- If the user wants the sound for a clip, make the clip first, then ask how long it is and use that as `seconds`.
- A voice recording of a real person is not something this skill does.