Skip to content
GenMedia Creative Studio

Chirp 3 HD Voices

The Chirp 3 HD workspace (/chirp-3hd) provides a dedicated studio for Google Cloud’s Chirp 3 HD Voices via the Cloud Text-to-Speech API (google.cloud.texttospeech_v1beta1), designed for enterprise narration, localization across 30+ languages, and exact phonetic pronunciation overrides.

Chirp 3 HD Interaction Architecture

Source Architecture

Section titled “Source Architecture”
Layer File Path Responsibility
Mesop Page & State pages/chirp_3hd.py (Chirp3hdState) Voice & locale selector, speaking rate slider, pause insertion helpers, custom pronunciation editor (IPA / X-SAMPA), and Gemini AI script generator.
Model Adapter models/chirp_3hd.py (synthesize_chirp_speech, list_chirp_voices) Calls TextToSpeechClient.synthesize_speech() with VoiceSelectionParams, AudioConfig, and CustomPronunciations, returning WAV bytes.

Key Capabilities & Controls

Section titled “Key Capabilities & Controls”

1. Dynamic Multi-Locale HD Voice Discovery

Section titled “1. Dynamic Multi-Locale HD Voice Discovery”
  • On workspace initialization, list_chirp_voices() queries TextToSpeechClient.list_voices() and filters for Chirp3-HD voice profiles (en-US-Chirp3-HD-Kore, es-ES-Chirp3-HD-Aoede, ja-JP-Chirp3-HD-Fenrir, etc.).
  • Users can filter voices by Language / Locale (BCP-47) and Voice Persona (Aoede, Charon, Fenrir, Kore, Leda, Orus, Puck, Zephyr).

2. Speaking Rate & Pause Controls

Section titled “2. Speaking Rate & Pause Controls”
  • Speaking Rate (speaking_rate): Adjustable from 0.25x to 2.0x normal cadence via texttospeech.AudioConfig(audio_encoding=LINEAR16, speaking_rate=...).
  • Pause Helpers: Quick-insert controls for natural pauses and punctuation pacing within long-form scripts.

3. Custom Phonetic Pronunciations (IPA & X-SAMPA)

Section titled “3. Custom Phonetic Pronunciations (IPA & X-SAMPA)”

For brand names, medical terminology, or regional proper nouns, users can define multiple Custom Pronunciation Rules (CustomPronunciationEntry):

  • Phrase: The exact substring in the input text to override.
  • Phonetic Encoding: Either International Phonetic Alphabet (PHONETIC_ENCODING_IPA) or Extended Speech Assessment Methods Phonetic Alphabet (PHONETIC_ENCODING_X_SAMPA).
  • Phonetic String: Passed to the Cloud TTS API via SynthesisInput.custom_pronunciations.

4. Integrated Acoustic Analysis & Library Persistence

Section titled “4. Integrated Acoustic Analysis & Library Persistence”
  • Synthesized LINEAR16 WAV audio is persisted to gs://<GENMEDIA_BUCKET> via store_to_gcs(), logged as a MediaItem in Firestore, and optionally analyzed with models/audio_analysis.py for pitch contour and voice quality inspection.