Everybody Develops Their Own "Tricks"
Like many of us who first dip our pinkie toe into the waters of AI Text-To-Art, I started with the most simple and cheapest models I could work with. In my case, it was Flux Schnell. This was just over a year ago, and for a while that was my go-to, and in fact, for a while the only model I'd use. Even once I had become a "pro" member on NC, I'd stick to the free model and spend my credits on enhancing and such.
But after a while, that got boring and limiting. And I tried a few other models to various levels of satisfaction. I had heard about loras but had no idea what they were. I did get a little bit of help from a couple of friends, some gentle and simple coaching, and suddenly virtually all I was doing was Flux Lora items. And I liked it a lot.
But as other models began to come online, I liked their base models better than I like Flux. And I had begun starting to play around with an idea that worked to a degree; used it on Flux Schnell for Daily Challenges as well as with some of the other models... an add-on to my prompt using some of the "trigger words" for the loras I liked.
I don't know if it's revolutionary. I don't know if it's Kosher. I don't know how anybody else feels, all I know is that words are words, and we use words to build the prompts we want.
For several months, virtually every creation I made had this as the final part of the prompt: "insanely detailed, vibrant colors, Hyperrealistic, intricately detailed, color depth, dramatic, 2/3 face angle, side light, colorful background LTp: 300mm on full frame photography balanced, vibrant colors photorealistic realistic photoreal real realistic Extremely detailed, Cinematic"
Yes, I know there is some redundancy in there. And occasionally I would add in a few more favorites, but these were the ones that were most often used.
I rarely use these now. The newer models we have access to seem to work as well or better without those trigger words. But it was a trick that worked for a while, and still does on rare occasion that I'm at a loss to get the look I'm trying for.
Comments (23)
I started on NC using Stable Diffusion 1.5 and SDXL BETA. When FLUX became the new standard, all the LoRAs I had used became useless. All the tricks I came up with also became useless. What a rude awakening. Now, I'm happy with Flux and have found some new tricks, primarily thanks to other artist (hi, Rob!)
Thanks for sharing. I think my path was similar. If I recall correctly, my original go to was Dream Shaper xl Lightening. Now I have a few models I will use, Seedream a favourite. Also have a few stock words that I tag on at the end, which are general for image quality, rather than look.
Yes, having friends who will share tips helps. As does the people who don't mind leaving their prompts open. I have found a lot of great ideas in other people's prompts. Usually just a line or a way of phrasing that I had not considered, that proves to be very, very useful!
I agree, it's like a "lite-lora" in effect, and of course I am borrowing from the work of the original lora-creator who determined that those trigger words had those effects, but then every word we use has been used before to communicate the meaning. I'm claiming invention of nothing in that regard. But you're also right in that it often gets the job done; perhaps not as blatant or "in your face" as a lora would... and sometimes that's a truly better result. Subtle is often closer to perfection that blatant LOL
I'd definitely look at sites with prompt enhancers. I use one I found on Perchance; it gives a variety of creative options and output styles. I don't think it will help copy any particular user's style, but you can definitly use the keywords double exposure and try other keys. I often will run a prompt I've created, enhance it, then create a version from both. More times than not I end up preferring the enhanced version
Sign in to read all 23 comments and add yours.