Judgment here is necessarily subjective. I will ignore small anomalies, but clear discrepancies from how reality works will get a "no". Technically believable images that violate the prompt will also fail: for instance, a video of a man walking behind a large tree and a differently-dressed man emerging will fail if the prompt was "a man on a walk, passing behind a sequoia". Likewise, the famous "horse riding an astronaut" prompt must properly depict the horse atop the astronaut.
Prompt-engineered versions of these prompts will not be considered a "pass"; these easily human-interpreted prompts must succeed without the AI being given any special assistance.
If I do not have access to Sora before end of year, or the price for said access does not go below $250 before end of year, all options will resolve N/A.
1,000
3.00Do clear discrepancies from how reality works but that are barely noticeable count?
Eg if a tree that's in the distant background of the video and thus very small and intentionally out of focus fades out of existence; i.e. such that you would never notice it when watching normally but also such that if you're looking for it it does clearly violate reality, would that count?
@Nat I'm not going to be going over the videos with a fine tooth comb; each of these represents a particular possible "point of failure". So if small things like that happen, but the core of the video works, it'll resolve "yes".
The die should not repeat pips until it gets back to the "start", which should be the same, but it's fine if it's not perfectly arranged. The tree doesn't have to exactly be a sequoia (though I doubt that will be an issue), just a suitably thick tree that someone can walk behind and come out the other side.