Sinapsis Zonos

Templates for advanced speech synthesis using Zonos

🐍 Installation • 🚀 Features • 📚 Usage example • 🌐 Webapp • 📙 Documentation • 🔍 License

This Sinapsis Zonos package provides a single template for integrating, configuring, and running text-to-speech (TTS) and voice cloning functionalities powered by Zonos. It supports multilingual speech, emotional modulation, and real-time audio generation.

🐍 Installation

Install using your preferred package manager. We strongly recommend using uv. To install uv, refer to the official documentation.

Install with uv:

  uv pip install sinapsis-zonos --extra-index-url https://pypi.sinapsis.tech

Or with raw pip:

  pip install sinapsis-zonos --extra-index-url https://pypi.sinapsis.tech

With uv:

  uv pip install sinapsis-zonos[all] --extra-index-url https://pypi.sinapsis.tech

Or with raw pip:

  pip install sinapsis-zonos[all] --extra-index-url https://pypi.sinapsis.tech

Zonos depends on the eSpeak library phonemization. The installation depends on your OS. For Linux:

apt install -y espeak-ng

🚀 Features

Templates Supported

• ZonosTTS: Template for converting text to speech or performing voice cloning based on the presence of an audio sample.​

  Attributes

  • cfg_scale(Optional): Controls randomness and creativity in speech generation (default: 2.0, range: 1.0–5.0). Higher values introduce more variation in speech output.
  • denoised_speaker(Optional): If True, applies denoising to the speaker embedding to reduce background noise (default: False).
  • dnsmos(Optional): Denoising strength for hybrid models (default: 4.0, range: 1.0–5.0).
  • emotions(Optional): Emotion configuration to fine-tune the emotional tone of the generated speech (default: {}). Accepts an Emotions object with weights for various emotions.
  • fmax(Optional): Maximum frequency cutoff in Hz for audio generation (default: 22050, range: 0–24000).
  • language(Optional): Language code used for synthesis (default: en-us)
  • model(Optional): The Zonos model identifier to use (default: Zyphra/Zonos-v0.1-transformer). Options: Zyphra/Zonos-v0.1-transformer and Zyphra/Zonos-v0.1-hybrid.
  • output_folder(Optional): The folder where generated audio files will be saved (default: SINAPSIS_CACHE_DIR/elevenlabs/ audios).
  • pitch_std(Optional): Standard deviation for pitch variation, which influences pitch naturalness (default: 20.0, range: 0–300).
  • prefix_audio(Optional): Path to an audio file used for prefix conditioning (e.g., whispering or prosody control) (default: None).
  • randomized_seed(Optional): If True, a random seed is used for each generation (default: True).
  • sampling_params(Optional): Controls sampling behavior for speech synthesis. Accepts a SamplingParams object with fields like top_p, top_k, min_p, linear, conf, and quad.
  • seed(Optional): Random seed used for deterministic generation. If randomized_seed is False, this value ensures repeatable output (default: 420).
  • speaker_audio(Optional): Path to a reference audio file used to extract speaker characteristics for voice cloning (default: None).
  • speaking_rate(Optional): Speaking rate in syllables per second (default: 15.0, range: 5–30).
  • unconditional_keys(Optional): A set of keys (e.g., {vqscore_8, dnsmos_ovrl}) that disable speaker conditioning when generating speech.
  • vq_score(Optional): VQ score threshold used by hybrid models to determine decoding style (default: 0.7, range: 0.5–0.8).

For example, for ZonosTTS use sinapsis info --example-template-config ZonosTTS to produce an example config like:

agent:
  name: my_test_agent
templates:
- template_name: InputTemplate
  class_name: InputTemplate
  attributes: {}
- template_name: ZonosTTS
  class_name: ZonosTTS
  template_input: InputTemplate
  attributes:
    cfg_scale: 2.0
    denoised_speaker: false
    dnsmos: 4.0
    emotions:
      happiness: 0
      sadness: 0
      disgust: 0
      fear: 0
      surprise: 0
      anger: 0
      other: 0
      neutral: 0
    fmax: 22050.0
    language: en-us
    model: Zyphra/Zonos-v0.1-transformer
    output_folder: ~/.cache/sinapsis/zonos/audios
    pitch_std: 20.0
    prefix_audio: null
    randomized_seed: true
    sampling_params:
      min_p: 0.0
      top_k: 0
      top_p: 0.0
      linear: 0.0
      conf: 0.0
      quad: 0.0
    seed: 420
    speaker_audio: null
    speaking_rate: 15.0
    unconditional_keys: !!set
      dnsmos_ovrl: null
      vqscore_8: null
    vq_score: 0.7

📚 Usage example

This example shows how to use the ZonosTTS template to convert text into speech. The generated audio is based on the input text and is saved locally as a file.

agent:
  name: text_to_speech
  description: text to speech agent using Zonos

templates:

- template_name: InputTemplate
  class_name: InputTemplate
  attributes: {}

- template_name: TextInput
  class_name: TextInput
  template_input: InputTemplate
  attributes:
    text:  This is a test of Sinapsis Zonos text-to-speech template.

- template_name: ZonosTTS
  class_name: ZonosTTS
  template_input: TextInput
  attributes:
    model: Zyphra/Zonos-v0.1-transformer
    language: en-us
    emotions:
      happiness: 0.3077
      sadness: 0.0256
      disgust: 0.0256
      fear: 0.0256
      surprise: 0.0256
      anger: 0.0256
      other: 0.2564
      neutral: 0.3077
    fmax: 24000
    pitch_std: 45.0
    speaking_rate: 15.0
    cfg_scale: 2.0
    sampling_params:
      linear: 0.5
      conf: 0.4
      quad: 0
    randomized_seed: True
    denoised_speaker: False
    unconditional_keys:
      - dnsmos_ovrl
      - vqscore_8

This configuration defines an agent and a sequence of templates for speech synthesis, using Zonos.

To run the config, use the CLI:

sinapsis run name_of_config.yml

🌐 Webapp

git clone git@github.com:Sinapsis-ai/sinapsis-speech.git
cd sinapsis-speech

docker compose -f docker/compose.yaml build

docker compose -f docker/compose_apps.yaml up -d sinapsis-zonos

docker logs -f sinapsis-zonos

Running on local URL:  http://127.0.0.1:7860

docker compose -f docker/compose_apps.yaml down

uv sync --frozen

uv pip install sinapsis-speech[all] --extra-index-url https://pypi.sinapsis.tech

uv run webapps/generic_tts_apps/zonos_tts_app.py

Running on local URL:  http://127.0.0.1:7860

📙 Documentation

Documentation is available on the sinapsis website

Tutorials for different projects within sinapsis are available at sinapsis tutorials page

🔍 License

This project is licensed under the AGPLv3 license, which encourages open collaboration and sharing. For more details, please refer to the LICENSE file.

For commercial use, please refer to our official Sinapsis website for information on obtaining a commercial license.