Clip and Audiogram Creation
Emotion Tags
normalslowcryingsleepysighchuckle
Episode Folder Organization
Guided Emotion and Intonation
Input Streaming for Lower Latency
LLM-based Customizability
Manual Work Reduction
Save up to 100% of manual work
Model Tokenizer Type
Non-streaming (CNN-based) tokenizer
Multiple User Types Supported
Content CreatorsPodcast ManagersProducers & StudiosAgencies
Open Source Release Planned
Orpheus Speech Models
Medium (3B)Small (1B)Tiny (400M)Nano (150M)
Pretrained and Finetuned Models
Pretrained modelsFinetuned models
Prompt-Based Content Tuning
Remove Silences and Fillers
Sample Finetuning Scripts
Sliding Window Detokenizer
Social Media Post Creation
Streaming Inference Speed
Faster than playback on A100 40GB for 3B model
Training Data Volume
100k+ hours of speech, billions of text tokens