YouTuber proves DLSS 5 is not AI slop

Nvidia announced DLSS 5 earlier this month with a demo that showed incredible detail being added in a handful of games. It became a controversial topic as despite the company’s explanations, the community believed it was AI slop being added to games. Now, a YouTuber has broken down the demo videos in an attempt to prove DLSS 5 does not generate new content and features, but only enhances what’s already there.
For context, the DLSS 5 preview showed comparisons of certain scenes from games like Resident Evil Requiem, Hogwarts Legacy, Starfield, and The Elder Scrolls 4 Oblivion Remastered with DLSS 5 on and off. The difference was astounding, with DLSS 5 seemingly adding detail that wasn’t in the original. This raised questions about artistic integrity and the final result not being true to what the artist intended.
With the analysis video (linked below), YouTuber WhizzDumbPlayz went through the demo scenes and dove into the primary concerns that were pointed out after the preview. Starting with Starfield, DLSS 5 seemed to add hair texture where there wasn't before and changing the size of the character’s nostril, giving them a different look. Upon slowing down the DLSS off footage, the YouTuber pointed out that the hair was already present, but rendering incorrectly, and that the nostril looked smaller than it actually was due to global illumination. Thus concluding that DLSS 5 only enhanced what was already there without adding or generating geometry.


Similarly, with EA Sports FC 26, the YouTuber pointed out that all the artifacting with DLSS 5 on was present in the footage without DLSS 5, and was caused by the motion blur implementation and frame generation adding more artifacts. Moving on to The Elder Scrolls 4 Oblivion Remastered, the base footage already had the ‘eyelids with eyes’ which was made more noticeable with DLSS 5. Same with the older woman in Hogwarts Legacy.
Arguably the most controversial footage was of Resident Evil Requiem showing Grace, the protagonist, standing still looking at the camera. The community at large believed that the it was a different character model that DLSS 5 was generating or showing since it added details that the original character model did not have. WhizzDumbPlayz noted that DLSS on and off shots were taken at slightly different angles and at different points in the idle animation. The DLSS on footage has her mouth slightly opened and head slightly tilted back. Aligning the footages shows a less drastic change between DLSS on and off. Furthermore, the YouTuber also points out that the original footage shows improper path traced lighting specifically around the right ear and hair area, which is then supposedly fixed with DLSS 5, making Grace look more realistic and, in turn, more AI generated.


WhizzDumbPlayz is confident that DLSS 5 is not adding detail or geometry but only enhancing what’s already in the game. Lastly, he points out that it is a tool that is in full control of the artist, and that the tool itself should not be blamed.
Source(s)
WhizzDumbPlayz on YouTube















