- Superhuman AI
- Posts
- A new model takes on Midjourney
A new model takes on Midjourney
ALSO: Creating product photo animations with AI
Zain Kahn
August 02, 2024
Read time: under 4 minutes
Welcome back, Superhuman
Germany’s Black Forest, the mystical region that inspired the Grimm brothers' stories, is lending its name to a new AI startup. And the company just unveiled its own enigmatic creation — not a fairytale, but a cutting-edge AI model.
Today’s Insights
An impressive new image generator launches from stealth
Tutorial: Creating product photo animations with AI
How AI can help prevent shark attacks
Everything else you should know today
5 new AI tools to boost your productivity
AI-Generated Images: Crazy frog
NEXT IN AI
A new text-to-image model soars to the top of the charts
Source: Black Forest Labs
An AI startup called Black Forest Labs just emerged from stealth — and immediately leapfrogged over the competition. Its suite of text-to-image models, known as FLUX 1, already outperforms rivals like Midjourney 6.0, DALL-E 3 HD, and Stable Diffusion 3-Ultra.
It leads across multiple factors, like image detail, scene complexity, and prompt adherence, according to the Elo rating system. And Black Forest says it has all of the aspect ratio and style options you’ve come to expect from today’s image generators, too.
Who’s behind it? Several of the startup's engineers come from Stability AI, the developer behind the influential image generator Stable Diffusion. That company ran into trouble earlier this year when its CEO, Emad Mostaque, was accused of mismanagement and eventually stepped down.
Black Forest is also backed by a who’s-who of AI industry insiders, like Y Combinator CEO Garry Tan and Timo Aila, a senior researcher at Nvidia — not to mention Andreessen Horowitz, which led the startup’s $31 million seed funding round.
How it's different: Black Forest says it combined several experimental training techniques to make FLUX 1 both faster and more accurate.
For one, it uses rotary positional embeddings — an approach that helps the model keep track of large sequences; each piece of data is assigned its own unique characteristics, which makes it easier for the LLM to tell them apart
It also uses a parallel diffusion transformer, which lets the model analyze multiple parts of a sequence simultaneously — speeding up the time it takes to transform visual noise into a cohesive image
What next? Black Forest says it’s planning to drop a state-of-the-art text-to-video model soon. If it’s anything like FLUX 1, other video-focused AI companies (including OpenAI, HeyGen, and Runway) might have their work cut out for them. In the meantime, you can try out FLUX 1 on cloud platforms like Fal and Replicate.
PRESENTED BY GAMMA
How to create beautiful presentations with AI
Go to the Gamma website and sign up.
Once you’re signed up, you’ll get an option to create slides by pasting text, uploading a file, or giving it a prompt.
You can pick any option you like. For this example, I’ll pick the prompting option.
On the next screen, pick how many slides you want, and the language, describe the content of the slides, and click generate.
You can also use its AI to find or generate images
Next, you can choose the design and colors you’d like for the presentation.
The app will create the presentation in a few seconds.
Edit the contents of the slides to suit your needs.
THE AI ACADEMY
How to create product photo animations with AI
Go to Mojo AI’s website and Sign up to get tokens.
Upload your product photo or company logo to convert it into animation.
Click on ‘Use this Image’, and on the next screen select any animation style from the list.
Wait a few minutes and you’ll get your perfect-looking animated logo or product photo.
You can use it as an Instagram story, reel, add to your website, or just add some text to make it more appealing and share it with your audience.
Want to master AI and future-proof your skills? Access 100+ courses and tutorials at our AI Academy
AI & THE OCEAN
How AI is helping prevent shark attacks
Would you feel safe swimming in waters monitored solely by an LLM? If so, you might be onto something. Researchers at the University of California, Santa Barbara designed a model that’s been fine-tuned to spot sharks in the water — and early tests suggest it might be more precise than humans.
Even with the help of aerial drones, people can only detect sharks with about 60% accuracy, according to CNN. The researchers’ model, SharkEye, can pick out most of the sharks humans see, but it’s also capable of spotting ones that elude us. That’s because it can peer deeper into the water, detecting creatures below the surface.
How it works: Like humans, SharkEye uses drones to find sharks in the water along SoCal’s Padaro Beach. If there’s a match, it’ll send out an alert to nearby lifeguards, parents, and business owners. For now, humans are still in control of the notification system. But as soon as next summer, the model could start taking on more responsibilities.
It could be good for sharks, too: As much as humans might feel threatened by sharks, the truth is that sharks are probably just as afraid of us. Human development and climate change have driven some species closer to shore, but better surveillance can help us find new ways to keep them safe.
THE AI ACADEMY
Make AI Your Superpower
Learn the AI skills that will help you get ahead in your career. Level up your command of the most powerful AI tools and workflows with:
100+ tutorials and workflows
New tutorials posted every day
Our Top AI tools + cheatsheets for work
A global community of professionals using AI
Built by our team of AI experts and practitioners to help you unlock your productivity at work. Readers of Superhuman AI get $100 off with the code SUPERHUMAN100.
AI & TECH NEWS
Everything else you need to know today
ChatGPT’s new voice model. Source: Swipe Insight
Power Hungry: On Meta’s earnings call this week, CEO Mark Zuckerberg said he predicts training Llama 4 will take about 10 times more computational power than Llama 3.
Dark Horse: An experimental version of Gemini 1.5 Pro that’s only available to developers just landed at number one on the Lmsys leaderboard — the first time Alphabet has claimed the top spot.
Long Time Coming: The European AI Act — the world’s first major AI legislation — officially went into effect Thursday, four years after it was conceived.
Karma Chameleon: Reddit has acquired the startup Memorable AI for roughly $40 million. The Massachusetts-based firm uses AI to optimize digital ads for different audiences.
Game Changer: New York’s Runway AI unveiled a new text-to-video model called Gen-3 Alpha Turbo that it says is significantly more efficient than its predecessor.
😁 One Fun Thing: Early users are having fun with ChatGPT’s long-awaited voice features. One user asked the model to sing “Happy Birthday” in a croaking frog’s voice — then in the style of a meowing cat, and finally, as a dog. Next, the user asked: “What if the dog was an opera singer?” Watch how ChatGPT responded here.
🧠 Brain Food: Taco Bell announced it’s rolling out AI voices across hundreds of drive-thru locations in 13 states. The chain said the new system will help with order accuracy and could reduce customers’ wait times.
PRODUCTIVITY
5 AI Tools to Supercharge Your Productivity
✅ Folderr: Create AI for any task, from a helpful chat assistant trained on your data to a powerful workflow automation for your business.
✅ Rome AI: An AI platform that creates podcasts for any topic you’re interested in by doing research, understanding subtopics, and crafting an episode you can listen to on the go.
✅ Not Diamond: Call the right model at the right time with the world’s most powerful AI model router.
✅ Amabay: Build a custom AI agent that can handle queries on your behalf.
✅ DeepKeys: An AI app that unlocks insights from your devices to enhance mental wellness and boost productivity.
PS: Want more? Check out our Top 100 AI Tools.
* indicates a promoted tool, if any
PROMPT OF THE DAY
Friday Funday - Emoji Party
Prompt: I want you to translate the sentences I wrote into emojis. I will write the sentence, and you will express it with emojis. I just want you to express it with emojis. I don’t want you to reply with anything but emojis. When I need to tell you something in English, I will do it by wrapping it in curly brackets like {like this}. My first sentence is “Hello, what is your profession?”Source: Beebom
AI-GENERATED IMAGES
Crazy Frog
Source: @_cyberone on Midjourney
Midjourney Prompt: A black and white photobooth film photostrip of a well-dressed frog wearing a Pez and sunglasses. The same well-dressed has different expressions in each photo
--style raw --ar 5:7 --v 6 --personalize asnzs4y
Acquire new customers and drive revenue by partnering with us
Superhuman is the world’s biggest AI newsletter for businesses and professionals with 600,000+ readers working at the world’s leading startups and enterprises. Companies like Amazon, Hubspot, and Salesforce feature their products in Superhuman. You can learn more about partnering with us here.