Would it be possible to add some new features?

[Alan-H1927]

Hi! Thank you for developing this amazing project. I’d like to kindly suggest two new features for future consideration:

1. Support for loading models locally: Currently, the project doesn’t seem to support loading models from a local path. It would be very helpful to add this capability, enabling users to work offline or use custom-trained models more easily.

2. Support for controllable generation parameters: At present, identical prompts (e.g., “Hammer slowly hitting the wooden table”) always produce the exact same output. It would be great to introduce generation parameters such as temperature or top-k, allowing users to adjust these settings and obtain varied, controllable results from the same prompt—enhancing both creativity and flexibility.

Thanks so much for your time and hard work!

[hungchiayu1]

Are you using different seed for inference?

[Alan-H1927]

Are you using different seed for inference?

I am currently not using any seeds. I infer by continuously changing the prompt words. However, I hope to have some parameters to control, rather than changing the prompt words.

[Alan-H1927]

Why is the seed parameter exposed in comfyUI, but not in other ways?