Skip to content

v1.2.0

Choose a tag to compare

View all tags
@sfox38 sfox38 released this 29 Mar 09:09
v1.2.0
47e3d94
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v1.2

New Feature: Random Voice

Configure a pool of language groups and the integration picks a different voice on every tts.speak call.

Setup: Press the 🎲 button next to Speak in the dashboard card, tick the language groups you want, press "Save and close". Then select 🎲 Random Voice from the language dropdown.

Automations: Pass voice: random in options:

options:
  voice: random

A fresh voice is picked on every call from both the dashboard card and automation callers. The pool persists across restarts and is independent of config entries.

Bug Fixes

Switching connection modes (proxy/direct) now works correctly. Disabling or deleting one connection mode while the other is active no longer destroys the shared dashboard entities (language, voice, device, message, speak button). Those entities now survive and transfer cleanly to the remaining connection entry.

Speak button no longer targets a stale TTS entity. After removing one connection mode, the button was staying locked to the old entity. It now checks loaded config entry states rather than potentially stale hass.states.

Message length. HA's text entity enforces a 255 character limit with a colour-coded live counter.

Language dropdown icon was rendering incorrectly. Replaced with a clean MDI translate glyph.

Random Voice state restore on restart was partially broken - the language dropdown restored correctly but the voice dropdown showed a real voice name instead of "Random Voice". Both now restore correctly.

Dashboard Card Changes

  • 🎲 dice button added alongside Speak button to open the Random Voice settings panel.
  • Speak button no longer full width.
  • Status feedback line removed as it was unnecessary.
  • Message textarea capped at 255 characters with a live counter.

Internal

  • Random voice pool stored via HA's Store class (.storage/tiktoktts_random_voices.json) - the correct HA-idiomatic approach for user preference data independent of config entries.
  • _RandomVoiceAwareCache dict subclass replaces HA's TTS SpeechManager.mem_cache to force cache misses for voice=random calls, ensuring a fresh voice is picked every time from any call source.
  • services.yaml added to suppress HA startup log warning about undescribed services.
  • Shared platforms (select, text, button) now tracked per config entry to prevent the "already been setup" error when switching between connection modes.

v1.1

New Features

  • New Voices: Now has over 100 unique voices, in multiple languages
  • Additional endpoints: For Direct API usage
  • Custom Dashboard Card: Use this card to easily preview voices or to send impromptu messages around your home.

v1.0

This is a modernized fork of the original philipp-luettecke/tiktoktts integration, updated for current Home Assistant versions with significant new features and bug fixes.

Important

If you are upgrading from the original integration, please read the Upgrading from the Original Integration section in the README before installing. A clean removal of the old integration is required.

Bug Fixes

  • Fixed PLATFORMS declaration: was a single value instead of a list, causing the integration to fail silently on modern HA versions
  • Fixed missing commas in SUPPORTED_VOICES: several singing/expressive voice entries were being silently concatenated into invalid strings by Python
  • Fixed default language and voice: changed from German (de / de_001) to English US (en_us / en_us_001)
  • Fixed broken GitHub URL in original en.json (pointed to wrong repository)

New Features

  • UI config flow: integration is now configured via Settings -> Devices & Services, no configuration.yaml editing required
  • Live connection test: endpoint and session ID are validated before saving, so failures are caught at setup time
  • Options flow: change endpoint, voice, or session ID at any time via the gear icon, changes take effect immediately without a restart
  • Direct TikTok API mode: optional alternative to the proxy that calls TikTok's internal API directly using a browser session cookie, with automatic fallback across 4 known regional endpoints
  • Automatic text chunking: direct mode splits long messages at sentence and word boundaries to respect TikTok's 200-character per-request limit
  • Retry logic: both proxy and direct modes retry failed requests up to 2 times with backoff before giving up
  • Stable entity ID: always tts.tiktoktts_proxy or tts.tiktoktts_direct, no longer a random UUID

Technical / Internal Changes

  • Migrated from legacy Provider base class to modern TextToSpeechEntity
  • Replaced deprecated async_timeout library with Python's built-in asyncio.timeout
  • All hardcoded strings (API paths, field names, status codes, retry config) moved to const.py
  • Config stored entirely in entry.data - single source of truth for both setup and options flows
  • Full developer documentation added to all files

Credits

Role Person / Project
Original integration philipp-luettecke/tiktoktts
Community TTS proxy Weilbyte/tiktok-tts
Voice list reference oscie57/tiktok-voice
Assets 2
Loading