X How to create books, images & videos in Turbo mode

Hello AInauts,

if you think the AI scene is on vacation ... think again, it certainly hasn't been the last few days! There were some really exciting updates that we are happy to report on - and have fun trying some of them out.

Let’s dive in, this is what we have in store for you today:

• 📚 How to create a book in one minute ...

• 👀 𝕏 with Grok-2: Uncensored pictures and a loose mouth

• 🎭 How to generate ultra-realistic images (and videos)!

• 📰 AI news quickie: The HAI highlights from the industry

📚 How to create a book in one minute ...

We have published many books by real authors on Amazon. And of course, since the advent of AI, we've been experimenting with what's possible (and what's not).

First we racked our brains on Google Sheets-based tools, endless ChatGPT conversations and specialized GPTs, and then we tried to get Gemini's massive context window to produce a long output. But it was never really usable ...

Sudowrite on the other hand was very good, but also involved a lot of manual effort. But now we’ve come across a new option that we would like to introduce to you today.

Infinite Bookshelf - fresh books in seconds

Developed on the basis of Meta's Llama3 models and the powerful Groq hardware, this app lets you create your book directly in the browser with a single prompt - faster than you can say "Simsalabim, write my book!".

1. Create an account with Groq (not to be confused with Grok, see below!) and get a free API key.

2. Open the app here and fill in the fields - the advanced version has even more parameters (Github from the creator Benjamin Klieger 🙏).

3. Simply enter a topic, click on "Generate" at the bottom - and the app generates complete chapters that logically build on each other

4. The app is particularly interesting for factual content, but fiction will also be better supported in the future.

5. After the generation, you can export the result as a text file or styled PDF.

6. Soon it will also be possible to save books directly to Google Drive and use notes as the basis for new projects.

And that's just the beginning, because new approaches are already on the horizon!

P.S.: The result is by no means ready for publication - but we think it's great for getting a comprehensive overview of any topic.

👀 𝕏 with Grok-2: Uncensored pictures and a loose mouth

You can think what you like about Elon (our index has been on the decline recently) ... but the richest man in the world is not only setting the pace with Tesla, SpaceX and Neuralink, he has now added a really solid AI to his X platform with Grok-2 from xAI.

The latest chatbot has a cheeky personality (we like it!) and can do more than just chat: it also generates controversial images of personalities, brands, etc., with virtually no filters ...

Risks and side effects: Images without filters and censorship

Images are therefore the most hotly debated new function of Grok-2. Almost every prompt is accepted, even controversial ones.

It is not even Grok's own model under the hood, but the new FLUX.1 model from Black Forest Labs, which is known for its photorealistic results (our article here). Elon has also announced that his company is already working on its own image AI.

This is a double-edged sword: on the one hand, you can unleash your creativity and create customized, (un)realistic images in a matter of seconds. On the other hand, such images lead to ... chaos, false reports, propaganda, slander, ...

Flashback: Remember the picture of the Pentagon explosion that caused a short-lived dip in the stock market? That will happen again, no half-hearted policy will prevent this...

And how good is the Grok-2 language model?

The predecessor Grok-1 was a big promise that disappointed in reality and couldn’t hold up in comparison with others. Even the open sourcing of the model could not hide this fact.

But Grok-2 is fun - and since the bot also has access to current information on X, it could be used in a variety of ways.

The latest Grok 2 model recently entered the LMSYS chatbot gladiator arena under the name "sus-column-r". There, it took one of the top spots on the podium, on par with OpenAI and Anthropic. Many even thought it might be a new ChatGPT model.

In addition to its big brother, there is also Grok-2 mini. This version was the first to be released on X. Both models will be available to companies via API in the coming weeks, and Grok-3 is due to be released this year.

Some data from European users is excluded from training - if you want to be on the safe side, you can actively switch this off.

What’s exciting in this context is that the field of language models is changing rapidly, and the best models on LMSYS have an ever-shorter half-life.

Our take: Promising, with potential for conflict

This time, Elon has produced more than just hot air with Grok-2. We got ourselves a premium account and have been playing around with it.

Grok-2 is not only smart and fun, but also up-to-date thanks to real-time information from X. Sure, the usual suspects like OpenAI and Anthropic aren't sleeping either. But with the pace that xAI is setting, it could get exciting. This thing has real potential.

Will the relatively uncensored images bring the regulatory authorities onto the scene? Certainly. But if AI-generated content is influencing public opinion and democracy, a broader discussion about it is definitely welcome.

Alright, if you want to check out Grok-2 for yourself, go to X and get a premium account for $8/month!

🎭 How to generate ultra-realistic images (and videos)!

Let's stay with visual content for a minute, because there have also been significant innovations with some other providers.

🖼️ Midjourney V6.1: New web editor for more control

The latest Midjourney update 6.1 brings a new tool that many designers will love: the web editor!

It simplifies the editing of images directly in the browser and integrates functions such as "Reframe", "Repaint" and "Vary Region". The new virtual brush gives you more control to get every detail of your image perfect.

Midjourney is also slowly moving away from its "Discord-Only" phase and opening up more and more to web use. This makes creative work much easier and more intuitive.

But while we are celebrating the new editor, the legal dispute with artists over alleged copyright infringements is entering the next round...

📸 Google Imagen 3: Not bad, pretty good!

Google has released version 3 of Imagen, which can also be used to generate detailed and photorealistic images with precise textures and lighting.

Google has published Imagen3 in the AI Test Kitchen... and the Internet is busy testing and comparing. Unfortunately, only available in the USA (or with VPN) for now.

So, that's the end of the topic of "pictures" (... Mystic will probably come in the next few days). These pictures are the perfect source material for videos. And someone has made a name for himself:

🎥 Gen-3 Alpha Turbo: 7x faster, but at half the price

Runway's Gen-3 Alpha Turbo has once again outperformed the competition.

Not only has the speed and cost been improved, but also the quality. The videos are more realistic, the movements smoother and the colors more vivid.

This is perfect for animating Flux images (at least until the new Flux video model is ready).

📰 AI news quickie: The HAI highlights from the industry

Finally, a few important updates at a glance:

• You are also using Claude, right? Why don't you install the app on your desktop, that will make access faster and easier.

• Btw, with the Claude for Sheets™ Prompt Engineering Template, you can generate answers directly in tables.

• And if you develop with Claude: the new prompt caching saves up to 90% costs and speeds up prompts by 85%. In other words: upload entire books or code repositories, cookbook here.

• A new AI scientist is able to independently generate research ideas, conduct experiments and write papers at a cost of less than 15 dollars! This brings us closer to automated research - video here.

• The GitHub open source project "Deep-Live-Cam" exchanges faces in real time in a live video stream - based on just one input image - tests here. Crazy, scary, brave new world.

• You can now test Hermes 3 for free on Lambda Chat or via API - the first fine-tuned Llama 3.1 405B model with improved reasoning and creative capabilities. Really good!

• Thanks to AI, your selfie could soon diagnose Marfan syndrome - with 98.5% accuracy and without a visit to the doctor. Will this also work for other rare diseases?

• OpenAI's new SWE-bench Verified tool measures how well AI solves real-world (software) problems - with human-validated tests that promise fair and accurate results.

• Cool new Hivemind app by grimoire creator Nick Dobos for iPhone/iPad/Mac. Create apps and publish them with one click!

That's it for today. We hope you liked it - see you next time!

Reto & Fabian from the AInauts

P.S.: Follow us on social media - that motivates us to keep going 😁!
Twitter, LinkedIn, Facebook, Insta, YouTube, TikTok

Your feedback is essential for us. We read EVERY comment and feedback, just respond to this email. Tell us what was (not) good and what is interesting for YOU.