The objective is to enhance the performance of the Iris - Face LQ-MQ model, especially in scenarios where faces appear at a distance or within wide shots. Currently, the model struggles with subtle distortions like blur or grain, leading to disfigured facial features in such instances. This inconsistency results in videos that look excellent in close-up shots but suffer from brief moments of disconfiguration when the shot widens or the subject moves farther away.
