inference.md

Diffsinger CLI Inference Notebook

---

This is simple notebook that aims to make CLI inference easier for everyone

Parameters

---

• DS_file = Path of the .ds file (You can generate one on OU by Export Project (Export Diffsinger Script v2)
• model_folder = Path of the acoustic model folder
• speaker_name = Name of the speaker to do inference
• lang = language
• gender = Gender parameter from -1 to 1
• output_path = Path to where to drop the generated WAV
• output_filename = Name of the generated WAV
• seed = seed used when rendering, leave it to the default -1 unless you are doing several tests and consistency is important
• key = Key transition of pitch
• render_steps = How many steps to render, more isnt always better!
• playback = Do you want to listen the result? Click this