Thaeyne
This video delves into the experiments that reveal how shifting the sequence of words can influence the resulting visual outcomes of Midjourney prompts. Witness the magic that transpires as linguistic elements shift, showcasing the nuanced dance between language and art. Join me in this illuminating exploration of word order and its influence on the artistic symphony of Midjourney.
Music generated with Mubert: https://mubert.com/render/pricing?via=thaeyne
All generated images created by Thaeyne with Midjourney: https://www.midjourney.com
Voiceover generated with: https://murf.ai
Talking images with: https://www.d-id.com/
Moving images with: https://kaiber.ai/
And more moving images with: https://runwayml.com/
And even more moving images with: https://www.decoherence.co/
0:00 Intro
2:47 Seeds and Speeds
4:04 No Space
5:32 Multiprompts
6:04 Word Order of Short Prompts
6:51 Word Order of Long Prompts
8:51 Outro
you rock thaeyne!! ππ½
Really love this channel itβs helped so much with my prompts and helped me expand on those
Don't use "in the style of SUBJECT …", try "SUBJECT inspired" or "SUBJECT inspired by" or "SUBJECT influenced by", these phases should make a difference.
Thank you for your videos. If possible a video describing ways to change the camera angle and view with prompting would be very helpful. πβοΈ
. I agree that in the style of does absolutely nothing other than adding noise, as does space between words. I got something for you to try, prompt something with a pronoun, a background and a style and then prompt a subject with plenty of visual description, then vary strong on the background image and add the subject as an image prompt, the image it generates will aways be interesting than anything you can get from straight promoting, even using multiprompting.
Great stuff, I would never have thought that "relaxed" vs "fast" would change the images. Does mid journey need to "think" longer..lol IMHO "in the style of" weakens whatever follows if what follows is not a specific art style. Thanks again for your great work.
With SDXL1 on Tensor Art, I used "In the style of" hundreds of time and never realized that the AI was ignoring me, even my spaces. Now I just put the name of the artist or style. I must say I'm disappointed. If (when) the AI becomes smarter, it will have to be able to sustain a conversation about the image. Ignoring words just won't be good enough. DALL-E3 already does this. MidJourney and Stable Diffusion must follow to keep up.
To me, it looks like midjourney favours what it knows best or maybe what it believes is more important for more of the audience? I guess weighting is able to force some words higher?
Thank you for very value vdo
Always sothorough greatideastopursue. Thankyou.
Will you be doing a series like this on DALLE with ChatGPT? It does some things amazingly well, like 3D models. It would be interesting to see your thoughts.
Aspect ratio often influences outcomes, especially with human figures. I hadn't thought that the speed would. Great stuff, as always. Cheers
my own word salad prompts are tweaked and refined every so often mostly by rearranging the order of the elements and at times choosing a different data training set protocol…if i obtain a few really nice images then i'll generate up to or over 100 images based on the same prompt and often isolate a given image to be evolved…there is a huge difference in style from my very first experimental images to what i'm producing now but that's to be expected