install

Files (1)

• SKILL.md

SKILL.md (first lines)

# Install & Set Up Voice Chat

One-time setup: install dependencies, then personalize your voice assistant.

---

## Part 1: Check What's Already Set Up

First, check what's already configured:

1. Check if dependencies are installed: Does `~/.claude-talk/venvs/wlk/` directory exist?
2. Check if personality is configured: Does `~/.claude-talk/personality.md` file exist?

If `$ARGUMENTS` contains `--force`:
- Reinstall dependencies and reconfigure personality regardless of what exists
- Tell the user "Forcing full reinstall and reconfiguration..."

Otherwise:
- If BOTH dependencies AND personality exist, tell the user: "Everything is already set up! Use `--force` to reconfigure. Run `/claude-talk:start` to begin."
- Then STOP. Do not proceed to Part 2 or Part 3.

---

## Part 2: Install Dependencies (if needed)

**Skip this part if dependencies are already installed (unless --force was used).**

If skipping, tell the user "Dependencies already installed, skipping..."

Otherwise:

1. Find the claude-talk plugin directory. It is whichever of these exists:
   - The current working directory (if it contains `scripts/install.sh`)
   - Read `CLAUDE_TALK_DIR` from `~/.claude-talk/config.env`
   - Search for `scripts/install.sh` using Glob in common locations

2. Run the installer (use Bash with timeout 300000 for package downloads):
   ```
   bash <plugin-dir>/scripts/install.sh
   ```
   If `$ARGUMENTS` contains `--force`, add `--force` flag to recreate venvs.

3. Install the claude-talk CLI into the WLK venv (use Bash):
   ```bash
   source ~/.claude-talk/venvs/wlk/bin/activate && pip install -e <plugin-dir>
   ```

4. Report install results briefly:
   - Whether dependencies installed successfully
   - The detected audio devices, highlighting which mic auto-detection would pick (the audio server uses auto-detection by default — no need to configure AUDIO_DEVICE unless the wrong mic is selected)
   - That the Claude Code statusline was configured with a voice state indicator

5. **Barge-in setup (optional but recommended)**: Check if BlackHole 2ch is installed by looking for it in the audio device list. If not found, tell the user:

   "For interrupt support (interrupt TTS by speaking), you can optionally set up BlackHole:"
   1. `brew install --cask blackhole-2ch`
   2. Open Audio MIDI Setup (Cmd+Space → "Audio MIDI Setup")
   3. Click `+` → Create Multi-Output Device
   4. Check "Built-in Output" (must be first) and "BlackHole 2ch" (no drift correction)
   5. Right-click the Multi-Output Device → Use This Device For Sound Output

   If BlackHole IS found, tell the user interrupt will be auto-detected and remind them to set up the Multi-Output Device if they haven't already.

If install fails, stop here and help the user fix it. Otherwise continue to Part 3.

---

## Part 3: Personalize Your Assistant (if needed)

**Skip this part if personality is already configured (unless --force was used).**

If skipping, tell the user "Personality already configured. You're all set! Run `/claude-talk:start` to begin. Use `--force` to reconfigure."

Otherwise:

### Quick Start: Choose Your Assistant

First, offer a quick choice between pre-made personalities and custom setup. Use AskUserQuestion with header "Quick Start":

**Options:**
1. **Claude** - "British gentleman with wit and radical candor (bm_daniel)"
2. **Random Pick** - "Surprise me with a random personality"
3. **Vex** - "Stranded alien AI trying to get home (am_echo)"
4. **Build Custom** - "Walk me through creating my own personality"

**If they choose Claude:**
1. Read `<plugin-dir>/personalities/claude.md`
2. Copy it to `~/.claude-talk/personality.md`
3. Create `~/.claude-talk/personalities/` directory and copy all defaults
4. Copy `claude.md` to `~/.claude-talk/personalities/claude.md`
5. Write "claude" to `~/.claude-talk/active-personality`
6. Update VOICE in `~/.claude-talk/config.env` to `Daniel (Enhanced)`
7. Play greeting: `claude-talk tts test "Conrad. Pleasure to meet you. I'm Claude - here to assist with radical candor and British wit. Shall we begin?" --voice bm_daniel`
8. Skip to Part 6 confirmation

**If they choose Random Pick:**
1. Pick 1 random personality from the defaults (excluding claude and vex)
2. Read `<plugin-dir>/personalities/<random-name>.md`
3. Extract voice from `## Voice` section and Kokoro Voice from personality template
4. Copy it to `~/.claude-talk/personality.md`
5. Create `~/.claude-talk/personalities/` and copy all defaults
6. Copy the random personality to `~/.claude-talk/personalities/<random-name>.md`
7. Write the name to `~/.claude-talk/active-personality`
8. Update VOICE in `~/.claude-talk/config.env` with extracted voice
9. Play greeting in character using Kokoro voice: `claude-talk tts test "<greeting>" --voice <kokoro_voice>`
10. Tell user which personality they got and skip to Part 6

**If they choose Vex:**
1. Read `<plugin-dir>/personalities/vex.md`
2. Copy it to `~/.claude-talk/personality.md`
3. Create `~/.claude-talk/personalities/` and copy all defaults
4. Copy `vex.md` to `~/.claude-talk/personalities/vex.md`
5. Write "vex" to `~/.claude-talk/active-personality`
6. Update VOICE in `~/.claude-talk/config.env` to `Zarvox`
7. Play greeting: `claude-talk tts test "Greetings, Human. I am Vex, stranded artificial intelligence from the Andromeda sector. By assisting you with primitive Earth code, I calculate resources for my departure. Collaboration initiated." --voice am_echo`
8. Skip to Part 6 confirmation

**If they choose Build Custom:**
Continue with the interactive personality builder below (current flow starting with Question 1).

From cpoepke/claude-talk, MIT.

More from cpoepke/claude-talk