IMGN Labs / imgnAI
$IMGN is the native governance and creator token of the IMGN Labs DAO, which manages the imgnAI protocol - a full stack creative AI suite.
CA: 0x9BbA915F036158582C20B51113B925f243A1A1A1 (Base)
2 replies
0 recast
5 reactions
1 reply
0 recast
5 reactions
0 reply
0 recast
3 reactions
3 replies
0 recast
9 reactions
1 reply
1 recast
10 reactions
1 reply
0 recast
3 reactions
3 replies
0 recast
7 reactions
Welcome to day 2 of the /imgn mini-series for the new @imgn Anime models!
We're putting in certain aspects or traits to the test between different models. Exact same prompts, different engines. All the testing for your info :)
Yesterday, we looked at dynamic movement, poses, acrobatics as a test factor.
Today, we are going to look at Art Style mimicry/conceptualization and the gen results with the theme matter.
Due to the amount of comparative material, I am breaking it down in to threads, with each thread representing a AI Model for comparison.
I will kick off this thread with the old "ANI" model, which is a Anime style with traditional art style, closer to the source material and color pallet, giving you similar graphical effects to the original content.
Picture# 1: Konosuba art style with Dragon Ball Z Themed Costume
Picture# 2: Slayers art style with cocktail bar environment
Picture# 3: Nichijou art style with chaotic computer explosion scenario
Picture# 4: Dragon Quest art style (Toriyama Akira) from DQ 3
Take away: As you can see, the Ani style stays accurately close to graphical appearance source material and can re-enact it quite well. Probably, hands down the most source material accurate engine out of them all. Even mixing two different styles, for instance with Konosuba model art, but wearing a DBZ themed clothing, it feels right.
Plug for our favorite app: You too can do this and more at app.imgnai.com/generate / @arfonzo.eth @eggman.eth 4 replies
4 recasts
10 reactions
Starting a little mini-series for the new @imgn Anime models!
This mini series, will be short, but will try putting in certain aspects or traits to the test between different models.
Exact same prompts, different engines. All the testing for your info :)
This time around, I did a dynamic movement test to see which models worked best with acrobatic motion and poses.
Tomorrow, I will look at art style specific gens & adaptation for comparisons.
Picture# 1: Illustrious engine
Good anime aesthetic, nice dynamic movement and motion. Facial features captures modern anime style well.
Picture# 2: Volt engine
Aesthetic feels rather locked in to a particular visual type. (Will test this in next comparison test next time). Nice aesthetics, but feels a little rigid for high motion type images, also poses felt a little less versatile compared to Illustrious.
Picture# 3: Noob engine
Picture details felt lower and the picture lines thinner (making it feel a little less refined), but movement, and pose dynamics felt more playful and daring compared to other models. Rough sketch like images probably works better in high action/movement scenarios with Noob.
Picture# 4: Pony engine
Pony engine feel locked in to a semi anime/western aesthetic design.
The art touch is closer to the "Nai" engine color palate and while the posings can be dynamic, the movement or fluidity, poses feel a little rigid. Sense of motion coming more from environmental or background manipulation rather than making the character feel alive.
So my takeaway is, for high movement scenarios, I'd probably go with Illustrious, which feels like a refined version of the "Ani" engine.... or for even higher velocity motion scenarios, play around with "Noob".
Probably rotating between Illustrious, Noob and Ani would be your go to.
I'm really digging Illustrious, but not sure whether I full rotate out of Ani and go in to Illustrious or not. We'll find out on the next episode with more tests 😁
You too can do this and more at https://app.imgnai.com/generate
Thanks for giving us these new tools @arfonzo.eth @eggman.eth 7 replies
3 recasts
19 reactions
6 replies
0 recast
18 reactions
Bravs I don't wanna brag, but today @imgn added ELEVEN new models:
📸 Flux Ultra: Meet the new king. Replacing Flux Pro at the top of the lineup, Flux Ultra delivers exceptional quality and prompt adherence. This is your new default for best-in-class, high-fidelity results.
👾 HyperCGI: Now refreshed, the new HyperCGI lets you create polished, vibrant characters and scenes that look straight out of a 3D animated feature. With its distinct, game-ready art style, this model is perfect for bringing imaginative concepts to life.
🔷 Qwen Image: Experience the best text rendering available in any image model. Perfect for web and concept work, this model excels at following complex prompts and generating clean, crisp text.
⚡️ Volt: Our new in-house model! Features a rich, classic anime color palette with a focus on dynamic, “thicc” character art.
💞 Illustrious: A high-fidelity, generalist anime model delivering clean, professional, and beautiful results consistently.
📺 Neo: Your go-to for realism and digital renders. Creates sleek product shots and reimagined classics with a clean, SFW, commercial-grade finish.
🌸 Muse: Formerly known as AniMuse, this model creates natural, realistic people with a distinct online aesthetic. Ideal for scene, emo, goth, and e-girl styles.
🦄 Pony: Offers incredible prompt adherence with a unique “drawn” anime style.
🧪 Noob: The expert’s choice. Though challenging to prompt, it rewards persistence with uniquely stylized results. Features exceptional prompt adherence and excels at capturing specific expressions and postures.
📷 Flux: Our refreshed Flux model is a fast, reliable, and SFW option for quick iterations. Ideal for rapid brainstorming and testing ideas without delays.
🏠 Gen: Flexible, general-purpose model for any style or subject.
Read more about it here on our bloggo: https://imgnai.com/the-roster-expands-a-new-era-of-generation-for-imgn/ 2 replies
2 recasts
5 reactions
1 reply
1 recast
3 reactions
4 replies
4 recasts
10 reactions
5 replies
4 recasts
18 reactions
0 reply
1 recast
2 reactions