Kling 4.0 vs 3.0: eight image-to-video tests, same image and prompt
The author ran eight 10-second 1080p image-to-video tests on Kling 4.0 and 3.0 with identical images and prompts, covering photoreal dialogue, 2D/3D animation, product, fast action, landscape, multi-shot and FPV, using first takes only and comparing audio.
Kling 4.0 vs Kling 3.0: same image, same prompt, one generation each. I ran eight image-to-video tests on both models at 10 seconds, 1080p: photoreal dialogue, 2D animation, 3D animation, a product shot, fast action, a landscape, a multi-shot scene and an FPV one-take. First result used, nothing re-rolled. Side by side below, sound compared in the thread. @Kling_ai 🧵
1. Photoreal with a spoken line. Both models spoke. Kling 3.0 added a word that was not in the prompt. Kling 4.0 kept the wording but slurred the middle of the line. Levels are matched in the cut so neither model sounds louder than the other. Kling 3.0 first, then Kling 4.0, each with its own sound. Prompt: The man on the left leans in and says, in a calm British accent, "You already know what I'm going to say." The other man lowers his cup and gives a small nod. Slow push-in on the two of them, rain running down the window, warm lamps steady. One continuous take, natural dialogue, cafe room tone.