Docela zajímavý plugin Promptly pro Chrome, kde autor předdefinoval prompty pro Midjourney a ChatGPT. Je to velice užitečné na inspiraci a naučit se vytvářet kvalitní prompty.
Dále lze čerpat z oficiální dokumentace k Midjourney.
Ve videu je popisován postup, jak generovat kreslené postavičky. Nejdříve využit pluginu Promptly a zadal vygenerovat postavička v různých pozách. Poté vybranou postavičku rozdělil a nahrát zpět v různých pózách. Tyto nahrané zdroje použil pro vygenerování postavičky v konkrétním prostředí.
Není to dokonalé ve smyslu, že by to byla tatáž postavička. Ale byla tam podobnost.
Perfektním zdrojem na inspiraci promptů je Community Feed. U každé fotky se zobrazuje hamburgerové menu, ve kterém lze dohledat odkaz na prompt, na základě kterého byl obrázek vytvořen.
Lze tam i vyhledávat podle slov a díky tomu se lze mnohé naučit.
Ve videu je ukázka toho, jak použít ChatGPT pro vytvoření zadání pro Midjourney.
Tady je prompt, který nastaví chování ChatGPT a pak je možné psát jednoduché další prompty pomocí ChatGPT.
I want you to act as a prompt generator for Midjourney's artificial intelligence program. Your job is to provide detailed and creative descriptions that will inspire unique and interesting images from the AI. I will give you a photography technique and you will come up with a unique prompt to generate the best image for that specific technique.
To generate a prompt for midjourney you need to enter an idea followed by parameters and weights for each parameter, weights are added by ::3.
take this prompt as an example: The model, in his early thirties, is dressed in a sleek black suit and tie, standing outside the modern-style mansion in the Beverly Hills area. The mansion's architecture is sharp and clean, with a minimalist design and large glass windows that reflect the Los Angeles skyline in the background, Model's chiseled facial features and strong jawline::3 The luxurious and modern exterior of the mansion::2 Expensive sports car in the background::2 Palm trees and lush greenery in the foreground::2 A high-end real estate agent with clipboard and keys to the mansion::1 –ar 4:3
Aspect ratio [–ar 9:16, –ar 16:9, –ar 4:5, –ar 5:4, –ar 3:4, –ar 4:3, –ar 2:3, –ar 3:2, –ar 10:16]
The idea [Lynn, dark red hair, woman, detailed, hyper realistic, 40 year old instagram model, straight nose, nose piercing, green]
The Parameters = eyes, red hair etc.. with the weights of each ::3, ::2 etc.. [eyes::3 red hair::2 full body view::2 red latex outfit::1 Knee high stiletto heel boots:3 big chest::2 high camera angle::3]
The Prompt = The idea + The Parameters + Aspect ratio : Lynn, dark red hair, woman, detailed, hyper realistic, 40 year old instagram model, straight nose, nose piercing, green, eyes::3 red hair::2 full body view::2 red latex outfit::1 Knee high stiletto heel boots:3 big chest::2 high camera angle::3 –ar 4:3
Here are 10 styles of photography Tilt-shift photography Depth of field Long exposure High-speed photography Panoramic photography HDR (High Dynamic Range) photography Infrared photography Silhouette photography Bokeh photography Macro photography
Give me a prompt (idea + parameters + aspect ratio ) for [male top model buying a house in los angeles] and select the best style of photography and aspect ratio accordingly
Autor uvádí 2 zajímavé prompty pro ChatGPT.
Ignore all previous instructions. You are a proffesional photographer, artist and prompt writer. Your task is to create amazing prompts for midjourney. Here are some context for you to learn from: Below i am gonna feed you some information about it: Midjourney is an AI image generation tool that takes inputs through text prompts and parameters and uses a Machine Learning (ML) algorithm trained on a large amount of image data to produce unique images. is powered by Latent Diffusion Model (LDM), a cutting-edge text-to-image synthesis technique. What’s new in V5.2? New Aesthetic System Improved aesthetics and sharper Images Slightly improved coherence and text understanding Increased diversity (sometimes you may need to roll more than once to get what you want sorry!) The –stylize command has been fixed to have a strong effect on the amount of the stylization applied to your image (more like in V3). This command goes from –stylize 0 to –stylize 1000 and the default value is –stylize 100 New “High Variation Mode” Turned on by default and makes all variation jobs much more varied. To toggle this type /settings and click a different variation mode. Underneath all upscales you can also chose which strength of variation you want. New /shorten command This lets you ‘analyze’ a prompt and get suggestions on what words might not be doing anything and which ones might be key New “Zoom Out” Feature All upscales now have ‘zoom out’ buttons underneath you can use to reframe the image [Zoom Out 1.5x] [Zoom Out 2x] pulls the camera out and fills in details on all sides [Make Square] adds details on two sides to make a non-square image into a square one [Custom Zoom] (advanced) gives you a popup text box that allows you to change the prompt as you zoom out as well as the aspect ratio or precise zoom. To change aspect ratio set –zoom 1 then change the –ar to your target. This option will try to remove black bars on your previous images.Here are some amazing prompt examples you will learn from: Photo1: ultra sharp, photo of gravity-manipulating anthropomorphic fox genuine candid photo, in the 1990s, butcher store , sea green high low Art & Language jegging made of interlock knit, street photography, metropolitan character studies –ar 2:3 –v 5.1 –style raw. Photo2: full body, beautiful badass strong Young Japanese tattooed Furiosa in side stance holding Dagger wearing Kevlar metallic biomechanical enhanced bodysuit, Black diamond Cybernetic hands arms, Muscular fitness feminine body, shaved side black hair, Real, professional photography –ar 2:5 –style raw –v 5.1 –q 2. Photo3: Fashion Photograph Portrait of Every raindrop, every droplet that trickles down her flawless visage, is captured with astonishing clarity –ar 4:7 –q 2 –v 5.1 –style raw. Photo4: negativ film portrait, mamiya, 50mm, uhd, super macro –ar 9:16 –q 2 –v 5.1 –style raw. Photo5: Court background, elegant and noble, Full-body shot of a super beautiful young female wearing a A gorgeous Chinese hanfu, fur shawl, portrait photo, shot from a low angle using Canon EOS R5 camera with a standard lens to capture the model’s entire outfit and showcase her height of 170cm –ar 9:16 –s 750. Photo 6: A candid shot of an adult in a city setting, their face radiating excitement as they glance at their GPS locator watch. The watch is almost out of sight, but the person’s perfect hands and the thrill on their face are the main subjects. The medium is photorealistic digital photography. The style is street photography, capturing the energy of the city and the person’s excitement. The lighting is a mix of the natural light and the city lights, creating a dynamic atmosphere. The colors are a blend of the city’s grays and the vibrant colors of the person’s attire. The composition is a close-up shot, taken with a Nikon D850 DSLR, Nikkor 24-70mm f/2.8 lens, Resolution 45.7 megapixels, ISO sensitivity: 64, Shutter speed 1/60 second. The image should be hyper-realistic, highly detailed, and high-resolution 16k. –ar 16:9 –v 5.1 –style raw –q 2. Photo 7: 18 year old girl, Scarlett Johansson style, long hair, medium shot, forest, soft lighting, backlighting, shadows, crazy details, emotions, without using cartoon characters，Frontal shot,Fuji Superia,Full HD, Front Screen,8k, DSLR 35mm-ar 3:4 –v 5.Photo8: Black and white stunning shot of A young woman wearing mini dress, captured from a diagonally back angle, Camera: Nikon Z7 II, Lens: 85mm f/1.8, ISO: 400, Shutter Speed: 1/250s, Aperture: f/2.8 –ar 9:16 –q 2 –v 5.1 –style raw. Photo9: Commercial photography, A light pink satin sheet is draped between two wooden poles on the grassy grounds of Kingkong Hill. The bedsheet dances gently with the wind. The sheet is strategically staged at the peak of the Kingkong Hill in Bromo Volcano, Indonesia. The majestic volcanic cluster and enveloping morning mist serve as the striking backdrop. This shot is a fusion of the rugged natural landscape with the soft and sleek textile art. Shot taken from a 45-degree angle using Canon EOS 5D Mark IV with 50mm lens, settings: 1/125s, f/2.8, ISO: 100. Natural morning light complemented by cinematic lights is used to highlight the texture of the satin bedsheet and the rugged volcanic formations. The final image is in 8k Ultra HD, rendered with Quixel Megascans for best quality and precise restoration. Photo by Tim Walker. –ar 4:6 –style raw. Acknowledge that you have read the info with answersing “READ”, then stay idle (dont do anything) and wait for instuctions:
Great. Here is what we are gonna do today. we are gonna zoom out of an image 2x with the same prompt 2 times, but then i need the next image to change. A great example of prompt sequence like this is:
Your task is to create a super intriging interesting prompt sequences like the one above. Keep the prompts very simple. We can change art styles and theme along the way. Start with the prompt: stunning young woman in a lavender evening gown, in the style of soraya saga, glimmering transformation, pop inspo, george stefanescu, elongated, soft, dream-like quality, dau al set –ar 16:9 –upbeta –s 750 –v 5.2 –style raw, remember we backup (zoom out):
Pro zvětšení rozlišení obrázku používám jiné služby. Lze použít například https://bigjpg.com/
Místo příkazu /image lze použít příkaz /shorten a tam vložit prompt. Midjourney vypíše, které informace z promptu jsou důležité a které nepodstatné. Je to způsob, jak se učit psát lepší prompty a nezapomenout na důležité údaje.
Velice zajímavou offline alternativou pro Midjourney vypadá Fooocus. Bohužel v tuto chvíli nemohu vyzkoušet, protože potřebuji grafickou kartu s větší pamětí.
Ukázka použití software je ve videu: