How to write good prompts? Prompt‑writing guides, or prompt‑assisting tools?
Thanks for sharing your model, but the videos I'm generating aren’t turning out as expected.
How to write good prompts? Prompt‑writing guides, or prompt‑assisting tools? any suggest?
or any work flow example
Use an AI LLM to create a detailed input reference image. I'm assuming you are referring to minimax h3 video prompts. Just ask an AI to generate your action prompt or even the complete prompt. There are alot of videos on you tube about creating good prompts for just about any model.
i dont see any difference
If you're using flux, it allows JSON. Paste this into flux for an example....
{
"composition" : {
"camera" : "half body view of female"
},
"scene": {
"description": "boxing gym",
"features" : [
"boxing ring with stained floor",
"weight racks along back wall",
"smokey atmosphere",
"harsh shafts of light from sun filtering through skylights"
]
},
"Female": {
"appearance": {
"body" : {
"breasts" : "large",
"hips" : "medium",
"legs" : "long and slender",
"skin" : "olive"
},
"face" : {
"eyes" : "emerald green with brown specles",
"lips" : "full"
},
"pose" : {
"description" : "statuesque",
"hands" : "taped with pink tape, fingers clenched in fists"
}
},
"hair" : {
"color" : "blonde",
"style" : "neat ponytail with slight curl",
},
"clothing" : {
"top" : "tight fitting pink sports bra",
"pants" : "pink Nike leggings"
}
}
}
JSON makes more sense with the tabs, but there are online tools for re-formatting it, or just ask an LLM to write you a quick Python script.
for minimax h3 use a grok minimax h3 prompt skill, thats what I do. then modify it as needed.
It's incredible, we have such a great mode of operation with JSON because it drives the language model to do exactly what we want. With these free flow text encoders we need a prompt enhancer to tell us how to talk to the prompt enhancer. Captain! Shields are at 5%...and...one more thing, the damage report machine is damaged. 😄