Skip to content

v0.6.1 — Les chats parlent

Choose a tag to compare

View all tags
@Gheop Gheop released this 11 Apr 17:15
· 92 commits to main since this release
v0.6.1
216fecf

The cats can finally speak. v0.6.1 closes the loop with v0.4.0 voice
input by adding voice output via Piper TTS, with a hybrid pipeline
that mixes real CC0 cat sound samples with French speech synthesis.

Highlights

  • 🎤 TTS hybrid pipeline — splits each AI response into alternating
    cat-sound + text chunks. `ronron Bonjour mon ami!` plays a real
    purr sample then a French voice saying "Bonjour mon ami", instead
    of a human pronouncing "ronron" phonetically.
  • 🐱 6 distinct cat voices — `fr_FR-upmc-medium` is multi-speaker
    (jessica, pierre); each catset character gets its own
    speaker_id × length_scale combo.
  • 🔊 Per-cat speaker toggle — click the icon in any chat bubble to
    mute that specific cat. Mid-sentence mute kills playback immediately.
  • 🔇 Cat sounds toggle in Settings → 'Voice output' for users who
    want text-only TTS.
  • 🎨 Hand-drawn pixel-art icons for the mic button, speaker, and
    transcription hourglass — matching CATAI's cream/brown theme.
  • 📅 Seasonal overlay first-launch-only — no more falling petals
    every time you restart CATAI.
  • 🗣️ Smarter system prompt — cats always respond with at least one
    full real sentence, no more onomatopoeia spam.
  • 🔧 Robustness fixes — TTS playback runs in isolated gst-launch
    subprocesses, mic press cancels in-flight TTS, chat bubble tail
    flips when the cat is near the screen top, text wraps around the
    speaker icon.

Install

```bash
pip install --user --upgrade 'catai-linux[voice]'
```

The `[voice]` extra now installs both `faster-whisper` (for voice
input) and `piper-tts` (for voice output). On first chat with TTS
enabled, CATAI downloads a 74 MB French voice model to
`/.cache/catai/piper/`.

See README for the full changelog.

Assets 2
Loading