This is a simplified guide to an AI model called Orpheus-3b-0.1-Ft maintained by Lucataco. If you like these kinds of analysis, you should join AImodels.fyi or follow us on Twitter.
orpheus-3b-0.1-ft is a text-to-speech (TTS) model that converts written text into natural-sounding speech. Built by lucataco, it represents the cutting edge in voice synthesis technology. This model joins other notable TTS implementations like whisperspeech-small and xtts-v2, offering distinct advantages in voice quality and emotional expression.
Model inputs and outputs
The model takes text input and generates high-quality 24kHz audio output using a two-stage architecture. The first stage converts text to tokens using a causal language model, while the second stage synthesizes audio using SNAC (Speech Neural Audio Codec) technology.
Inputs
- Text: The written content to convert to speech
- Voice Selection: Choice between tara, dan, josh, or emma voices
- Temperature: Controls generation randomness (0.1-1.5)
- Top P: Nucleus sampling parameter (0.1-1.0)
- Repetition Penalty: Prevents repetitive patterns (1.0-2.0)
- Max New Tokens: Limits generation length (100-2000)
Outputs
- Audio File: WAV format audio at 24kHz sample rate
Capabilities
The system produces clear, expressive s...
Top comments (0)