Early access demo shows Kling 4.0 video model following prompts in every detail
aziz4ai · x · 2026-09-29
Creator aziz4ai shared early-access results from the Kling 4.0 video model, claiming outputs follow every detail of the prompt verbatim.
His tips:
- Generate 3 results per run and pick the best
- Focus prompts on emotions, details, realism, camera moves, character motion, and physics
- The results are convincing enough that viewers might ask what movie it's from
A third-party hands-on showcase of the new Kling video generation capability, not an official release.
Related event: Early Kling 4.0 Flash tests show strong prompt understanding(2 posts)→
More from Multimodal
- Adobe's AI Relight feature rolls out in Photoshop beta — tisch_eins · 2026-09-29
- One creator and AI agents made a cyberpunk fashion film in 19 hours with ~50 prompts — full blueprint public — adrianscottcom · 2026-09-29
- AI-generated action figure fight: Mega Muscle Martha VS MogMog — KloneStudios · 2026-09-29
- Testing Qwen Image 2.1 vs FLUX Klein 2 vs Krea 2 vs Z Image in ComfyUI — aigeek1 · 2026-09-29
- Minimax H3 Video Imagines Superman as a Saiyan in Fan-Made Crossover — rogerro007 · 2026-09-29
- New video benchmark ranks 15 text-to-video models — chancemehmu · 2026-09-29