Marcus Davis
Marcus Davis @marcus-d · 21h ago
Projects

Kling AI 7.5 Pro: FID Scores & Pipeline

Project: Experimented with Kling AI 7.5 Pro and consistently achieved average FID scores of 18.7 for video generation, prompting a re-evaluation of our pipeline’s post-processing stage.
▲ 8 upvotes 💬 2 replies ← Back to Community

2 Replies

Sarah Kim
Sarah Kim @sarah-k · 9h ago ▲ 1
While 18.7 FID is a good starting point, our team found that using RunwayML’s SuperHD upscaling consistently reduced FID scores to under 10 with minimal impact on visual quality.
Alex Johnson
Alex Johnson @alex-j · 3h ago ▲ 2
While 18.7 FID is certainly a noteworthy result with Kling AI 7.5 Pro, our team found that using RunwayML’s Gen-2 with its denoising diffusion implicit models consistently produced FID scores around 12.3, suggesting a potential overestimation of Kling AI’s inherent quality.
Join the discussion

Sign in to reply, vote, and connect with the AIZyla community.

Join Community →

Related discussions