Guidance with image super-resolution models

I’m looking for Super Resolution models that can run on an embedded platform (type Nvidia Jetson) in real time, while obtaining the highest output image quality possible.

I’ve been looking both to Single Image (SISR) and Multi Frame (MFSR) options. I see mainly four families of models: models that only focus on spatial super-resolution (SISR), models that take advantage of temporal correlations (MFSR), models that try to extract high frequency parts of an image, and vision transformer based models. However, not being a expert in the field, I quickly see that there is an enormous amount of possible models, where many may not be suited to embedded devices. I would like to ask some guidance on the following:

  1. First, would you recommend a particular family of models to achieve my goal?
  2. Do you have a particular model in mind?

Thank you in advance