- The AI Stuff Creators and Artists Should Know
- Posts
- What's wrong with Midjourney 6.1?! πΌοΈ
What's wrong with Midjourney 6.1?! πΌοΈ
Hey, Creatives!π
Check the latest advancements in AI image generators and elevate your creative projects to new heights.
Topics of the week:
πΌοΈ Using Image-to-Video in Gen-3 Alpha
What happened to Midjourney 6.1?? π
π AI Image Generators Review
Creative Vision with Generative AI π₯β¨
π C.A.S.H. Camp
πΌοΈ Using Image-to-Video in Gen-3 Alpha
π Link: Using Image-to-Video in Gen-3 Alpha
This guide from RunwayML explores the new Gen-3 Alpha feature, which allows you to convert static images into dynamic videos seamlessly.
Sample Prompt
A cinematic documentary hand held close up of a woman standing in a busy Italian plaza smirking to herself, the background soft and out of focus, diffused overhead lighting. Her skin has freckles and small creases, her hair is down and a bit messy. Muted colors, diffused cinematic lighting, cool color grade.
View more of Runwayβs use cases here.
β¨ Personal Thoughts
I think Runway Gen-3 Alphaβs text-to-video quality has improved in terms of details and fidelity. However, I noticed some issues while testing it with human figures in clips with more motion. The faces and bodies tend to morph and distort, indicating that it still struggles with generating realistic, consistent human visuals. So, itβs somewhat limited when it comes to maintaining realism and consistency.
π₯ Experimentation Clips:
Clip 1
Requested more human movement and motion.
Observations: Significant morphing and body distortions; the womanβs face changed from Asian to Caucasian.
Morphing and Body Distortion
Clip 2
Requested minimal motion, just a simple camera zoom.
Observations: More consistency, but subtle changes are still visible on the woman's face.
Camera Zoom
πWhat happened to Midjourney 6.1??
Midjourney 6.1 is out, but we have mixed feelings about it. While there are some improvements, we've noticed a significant increase in the generation of naked women, and the hands seem to be looking worse compared to the previous version.
Read full the article here.
π AI Image Generators Review
Discover how Flux, the latest open-source AI image generator, outperforms top models like Midjourney, SD3, and Auraflow with its advanced 12-billion parameter system, offering unparalleled realism, versatility, and customization. Explore detailed comparisons and see why Flux sets a new standard in AI image synthesis
Comparative Insights
Feature | Flux AI | Midjourney | SD3 | AuraFlow |
---|---|---|---|---|
Artistic Style | π Versatile in generating different styles | π¨ Known for its artistic style | π§ Advanced customization options | πΌοΈ Produces realistic images |
Ease of Use | π Intuitive and user-friendly | π User-friendly interface | π― More control over final output | π High detail and accuracy |
Speed | π Fast image generation | β‘ Produces visually appealing images quickly | π οΈDesigned for fine-tuning with precision | π Ideal for high realism tasks |
Target Users | π Beginners to professionals | π‘ Artists and designers needing inspiration | ποΈ Professionals needing detailed adjustments | β High-quality, true-to-life visuals |
Customization | ποΈ Offers a variety of settings | π¨ Limited customization | βοΈ Extensive customization options | π§ Specific adjustments for realism |
Comparative Example
Realism
FLUX
Flux AI:
π Captures a bustling city street at night with realistic neon signs and wet pavement reflections.
π₯ People and cars are well-represented.
π "Decrypt" sign is prominently displayed.
π Ranked highest for detail and lighting.
AURAFLOW
Auraflow:
π¨ Vibrant neon lighting and realistic reflections create a bustling atmosphere.
π Street vendor visible but cartoonish.
π Image is over-saturated with blurry neon signs and distorted perspective.
π Ranked lowest due to deviation from hyper-realistic style and perspective issues.
SD3 MEDIUM
SD3 Medium:
π¦ Balanced composition with realistic lighting and reflections.
ποΈ Prominent "Decrypt" sign and lively atmosphere.
π Some unrealistic elements (people walking on the street, expanding sidewalk).
π₯ Ranked second for overall balance and composition.
Comparison of Flux and Midjourney for Realism:
Midjourney | Flux |
Prompt:
A black and white photo of a woman with long straight hair, wearing an all-black outfit that accentuates her curves, sitting on the floor in front of a modern sofa.
Midjourney:
π Features a dynamic, crouched pose capturing high-fashion essence.
π¨ High precision in hair, facial features, and clothing rendering.
ποΈ Unnatural elements: hand-foot mix, disappearing right leg, left foot resembling a hand.
π₯ Ranked second due to unnatural pose and inaccuracies in body representation.
Flux:
π§ Captures a natural, relaxed pose with legs crossed.
π High precision in rendering facial features, hair, and clothing.
π‘ Soft, diffused lighting with gentle shadows and highlights.
𦡠Minor flaw: additional leg, fixable with inpainting or Photoshop.
π Ranked highest for natural pose, contextual background, and detailed rendering.
Overall Ranking:
Flux: Most accurate in terms of morphology and overall scene realism.
Midjourney: Dynamic pose and detail, but lacks naturalism and has body representation issues.
Creative Vision with Generative AI π₯β¨
I recently attended @theand.foundation event and showcased a video created with @luma_ai, using images from Midjourney and edited with CapCut. From soundtracks to effects, everything was crafted with these tools, making the process both fun and efficient.
Disclaimer: These clips arenβt flawless. They demonstrate some limitations of current generative AI, like human body distortions. Sharing these videos helps keep everyone informed about the state-of-the-art.
π C.A.S.H. Camp
Here is an amazing work from our Camper Lucy Schoenfeld.
Testimonial from Lucy
Check more of her work on her Instagram @lucyschoenfeld
π Next Cohort Starts: This Friday, August 9th
Our C.A.S.H. Camp is starting in just a few days, but you can still join us on time! Whether you're an AI newbie or a seasoned pro, our C.A.S.H. Camp has just been upgraded!
Added bonus lessons:
AI-driven ( product ) videos
The Midjourney Web UI
Want to learn about creating a video clip using just 2 still images?
Hereβs a sample of our AI-driven product video clip:
Images: |
Final Clip
Future-proof your skills
Upscale your projects
Stay ahead in the creative industry
π Ready to join? Don't miss the chance to elevate your creative skills with AI. If August doesn't work, more cohorts are coming up!
That's all for this week!
β¨ Community Corner: We Want to Hear from You! β¨
Got a burning question or a puzzle for us to solve? π€ Submit your question here, and letβs get the conversation started! Weβre excited to explore every question and share the knowledge in our upcoming editions.
π Stay in the Loop!
Did you know that you can now read all our newsletters about The AI Stuff Creators and Artist Should Know that you might have missed.
Curios about C.A.S.H Camp Click here Book a free 20-minute discovery call with me. It's the perfect no-strings-attached way to explore if this is the path for you.
Don't forget to follow us on Instagram πΈ: @clairexue and @moodelier. Stay connected with us for all the latest and greatest! π¨π©βπ¨
Stay in the creative mood and harness the power of AI,
Moodelier and Claire πβ¨