catherinebazinaw
Catherine Bazinaw
Illustration and Graphic Design
227 posts
Don't wanna be here? Send us removal request.
Last Seen Blogs
lone-wolf11
WAZZZAPPP!!!
whereismh370-blog
Where Is Malaysia Airlines Flight MH370?
melbournedream
Melbourne
mordekayverigbioyunlari-blog
Cartoon Network Kahramanları Mordekay ve Rigbi
nanticchia-blog
I AM NOT HERE
Text
Training an AI Model
I decided to experiment. I'm a digital illustrator, graphic designer and artist so part of what makes me curious about AI is how I can integrate my style to expand my creative freedom.
I've been heavily inspired by the Odd Birds AI Project in which, Swedish artist Arvid Tappert has been exploring how to create an AI model, which allows him to generate 'odd birds' in his animation style, adding texture drawings and animations, frame-by-frame animation and walking cycles.
I'm very much in the early learning phases, exploring the capabilities of Dall-E, Pika Labs and Runway ML and how far I can stretch these platforms.
I have jived pretty well with Dall-E - I prefer its ethically trained models, and the way it enables me to flex my creative writing to create the imagery I want. I grew particularly interested in Renaissance inspired artworks that incorporated unusual fantasy elements and storytelling, creating a few really interesting works:
Once I had become familiar with the quirks of Dall-E and its prompting style, I started to prompt my poetry and really enjoyed the creative interpretations of my work:
From here, I started to bring it into my world, incorporating a greater level of Photoshop editing and photography to create a self portrait, and a portrait of my husband. The aim here to try and create consistent environments to elicit the illusion we are in the same space. This was achieved with artist's tools, the light blue porch, white frames, wood, and wind chimes.
These were particularly special to me, and inspired me to move into portraits of my beloved pets in much the same way:
With my use of Dall-E established, I continued to work on additional art pieces, tightening the skill and establishing a style within the AI realm. Next, it was time to explore how far I could take synthography. I was particularly interested in retro, sci-fi aesthetics in worlds that were intricate and impossible. This gave way to the Blown Glass Desert series:
From here, with talk of Sora, I began to explore ai video methods coming across the work of Moss Carpet and voidstomper (formerly thephantom_ai), and generation platforms Pika Labs and Runway ML via a reddit sub r/aivideo. This sparked a new wave of excitement, and I began inputting prompts I'd used in previous Dall-E image generation, as well as existing imagery that I wanted to animate, learning the quirks of these platforms to generate some really cool results!
My first collection created entirely with Pika Labs was inspired by my Blown Glass Desert series, building on the atmospheric 1970s sci-fi aesthetic along with some of my earlier prompts which combined different animals, and human/animal characteristics to create haunting and unsettling visuals. Coupled with some CapCut stock audio that was equally unusual, I built my first visual experience from my favourite generations:
youtube
In an effort to understand the limitations, and peculiarities of Pika, I started to venture out exploring the stylisation of the platform with a continued interest in 1970s arthouse, sci-fi and dark fantasy scenes, aesthetics and creatures. More to come in future posts.
0 notes
Text
In the Wildflowers
Digital illustration, ai generation. April 2024
A combination of ai components and original illustrations, these works were created using a model trained exclusively on my own bank of illustrations to produce work that follows my own aesthetic and style. The model is dubbed the CBModel.
0 notes
Text
Bing's Pixar AI Trend
Recently, I've been messing around with a Bing AI trend - Disney Pixar posters. It's been a lot of fun. It started with the dog trend for Dogstagrams, but it evolved from there. It took some getting used to when writing prompts - AI is somewhat temperamental when it comes to interpretation of certain words and phrasings. It certainly struggled when trying to ask it to create a single image with all of my pets! The image of me and my four animals below required a little careful photoshopping to add Binx, and you can see that Gizmo has long fur instead of short. But Star and Snowdrop were a pretty accurate likeness.
The skiing shots were just a bit of fun - what would make an enticing movie poster? Action is required! (I've never skied). And of course with halloween round the corner, who could resist?
Once I got comfortable with the quirks of writing prompts, I started exploring how I could use the trend to recreate powerful moments and memories. How I could bring together the things that I loved into one space. A couple of examples include an early memory of my husband out in his yard chopping trees, building bonfires and dancing by moonlight. The second being our wedding day - this actually presented some challenges. My dress was a blue silk tea dress, and my husband had a short, but neat mohawk - kind of military in style. But these were nearly impossible to recreate in AI. But the vision of our wedding is here. Scarlet Begonias in my hair and on his lapel, the Grateful Dead song by the same name playing in the background (as represented here by the poster/title), an art gallery setting, the icy blue of my dress and a vintage air about it.
And from there, I started to explore how I could use the trend to explore the emotions I have been feeling and dealing with, and the emotions I wanted to feel and the peace I was longing for and struggling to find. And from that I created the following series - and it's on going. Kind of like a diary, I made a conscious effort to consider how I was feeling and what I needed, or wanted, and how that could be visualised. You can see an exploration of my personal passions and emotions of the past week. The search for joy, inspiration, calm, peace, and the pain of heartbreak. The deeper I go, the more this character unfolds, and honestly, I feel super inspired by her.
None of the above images came out perfectly - each one required post edits and tweaks, for which I used Photoshop, and even then the likeness of each character could be improved. But it's really exciting to see how these trends can help inspire creativity in different ways. And it's been a lot of fun considering how my words can create solid, wholseome and somewhat complete images to evoke feeling and humanity.
I'd like to take this a step further now. I've been taking part in a Photoshop animation course, but I've been struggling to latch on to a visual concept that really works for me. These images might have just hit the nail on the head, so I'd like to take these and create my own versions in my own style that I can translate into animations.
So, watch this space as always. And embrace all new tools and innovations. You never know how they might carry your creativity forward.
...
If you'd like to take part in the Bing trend and try it out for yourself, visit Bing's Text to Image page, enter your prompt and log in to your Microsoft account - or create one for free. If you want to join the Disney Pixar trend, simply include the prompt 'Disney Pixar-style movie poster' (drop Disney if it blocks you) and if you want to include a movie title, add 'with the title "Title Name"...'. In case you're not into Pixar - it works with other themes too! Like 90s side-scrolling bit-art! It's really down to where your imagination takes you!
1 note
·
View note
