Video wird geladen...
Video konnte nicht geladen werden
GPT-6 AstraでBlenderを操作し、プリビズからAI動画を作る流れをよく見かけるようになりましたね! でも、風景の世界観を再現するには、3Dをどこまで作り込めばいいのか? 同じ街並みを3段階の詳細度で作り、元画像の質感や雰囲気がどこまで反映されるか比較しました! 詳細は引用の記事にまとめています👇
505,610 Aufrufe • vor 5 Tagen •via X (Twitter)
17 Kommentare

Thank you so much for sharing

I always go for having more details in Blender. It should prevent certain slippages and keep the scenes consistent.

TL;DR: low res grey-boxing is all you need It’s the first side-by-side comparison I’ve seen!

Entirely depends on if you are recreating a real environment versus a generative one. The base materials needed for full generative environments is incredibly low. However, if you are wanting consistency in material elements then there is a sliding scale of detail needed in the blender file.

Actually like the low one !

Interesting test thank you for sharing

This is great 🔥

Thanks for sharing, i've found through my testing that the low detailed blocks works very well. Especially if you considered the speed at which it takes to create the previz for a scene you're working on

Is this utilizing Triplo, or just Astra + Blender?

AstraとBlenderのみです!

…

Knowing how rough previz can be saves enormous time.

私もぜひテストしてみたいです!もし可能でしたら、このモデルを共有していただけますか?

Three levels of detail compared, genuinely useful test.

low detail still looking clean ngl

看起来不一定越细越好,氛围和镜头语言反而更关键

プリビズの作り込み度合いで質感の反映がここまで変わるのは面白いですね。世界観の再現には土台の3Dがやっぱり効いてくるんだなと勉強になりました
