open-mmlab/Amphion ? reverse-engineered prompt

Build me an open source toolkit for audio, music, and speech generation that makes it easy to try and compare different models in one place.

I want it to support text to speech, voice conversion, singing voice synthesis, singing voice conversion, text to audio, and a few vocoder and evaluation tools too. Make it feel like a research friendly project, with clear configs, reusable training and inference code, and enough examples that someone new can get started without guessing. It should also include simple visualizations of the model flow or architecture so it is easier to understand what is happening.

Please organize it so people can run different experiments, preprocess data, train models, generate audio, and evaluate results. If you need current setup details or model ideas, look up the latest docs online. Keep it practical, polished, and focused on reproducible audio generation research.