>>29773909
'--mode=extremely-spicy-or-crazy' is the full tag and its the tag suffixed to your prompt when sent to the video generator with the others being simply mode mode=normal and mode =custom and --mode=extremely-crazy
You can swap them out in the payload and despite on groks side it doesnt change the actual mode in the Json sse stream i have personally verified after many many hours across multiple version of the video generator confirmed it changed the upsampled prompt created by grok.
They no longer expose the upsampled prompt but the last version that did and what i can confirm as still being used now is a pretty detailed json. and between that version adn the previous that was simply a more detailed prompt based on the image and your own custom text the difference between --Mode=custom and --mode=extremely-spicy-or-crazy being in the payload was night and day.
A tip i could give until my browser extension is stable enough to be released is to either put your actionable prompt as text on the image OR in your custom prompt "text_elements": "Red bold title: Enter your prompt here of actions". using text elements will at frame one of the video most of the time display your text your wrote and ive used ti alot to bypass some moderation due to i assume the upsampler taking it in as part of the json to be displayed only as a title but the video model they use takes text on image to influence the actual video idk.
Also B64 is another good one. None of these methods are 100% going to work all the time but it ups your hit rate.