higgs-v2-llm

Higgs Audio V2 LLM Model

This is the Higgs Audio V2 3B parameter text-to-speech generation model.

Model Description

The Higgs Audio V2 LLM is an end-to-end multimodal model capable of understanding and generating high-quality text-to-speech audio.

Model Details

• Model Type: Text-to-Speech Generation Model
• Parameters: 3B
• Framework: PyTorch
• Architecture: Transformer-based multimodal architecture

Usage

from higgs_audio.serve.serve_engine import HiggsAudioServeEngine

engine = HiggsAudioServeEngine(
    model_name_or_path="Toowired/higgs-v2-llm",
    audio_tokenizer_name_or_path="Toowired/higgs-v2-tokenizer",
    device="cuda"
)

Model Architecture

• Text encoder for understanding input text
• Audio decoder for generating speech
• Multimodal fusion layers
• 3B parameters for high-quality generation

Intended Use

• Text-to-speech synthesis
• Voice generation applications
• Audio content creation
• Research in speech synthesis

Limitations

• Requires significant computational resources
• Optimized for English text
• May require fine-tuning for specific voices

License

Please check the original model repository for license information.

Contact

For questions about this model, please contact the repository owner.