🎥 Two new ingenious AI video models

AI-HOI AInauts,

Welcome to the latest issue of your favorite newsletter. Just before Christmas, some more news and updates from the two big players, OpenAI and Google. And, as always, plenty to try out.

Here's what we have in store for you today:

• 🎥 Two new ingenious AI video models

• 🎆 Google is also stepping up its game with AI images (try it out for free)

• 💻 New ChatGPT Projects make the tool 3x more useful

Let's go!

🎥 Two new ingenious AI video models

Let's kick off the week with two banging updates - one of which you can try out right now!

There's a lot going on in AI video again, and we have to expand on our article from last week.

Two new versions of AI video models have been released:

1) Google launches new Veo 2 model - is it better than Sora?

Everyone has to decide for themselves which models feels better. But Veo 2 is definitely in the top league:

• Clips up to 8 seconds long

• Particularly realistic presentation

• Fewer hallucinations

Unfortunately, we haven't been able to test it yet ourselves, but we are on the waiting list. But from what we've seen, it really does seem to be much better when it comes to generating realistic videos with physical elements.

The models often hallucinate fingers, or the movements no longer have anything to do with our physical laws. Take a look at the following comparison:

Blaine compares all the major models. And with a very challenging task: cutting steak.

Pay attention to hands, cutting movements and the interpretation of the perfect steak. This is where we think Veo 2 is by far the best.

You want to try Veo 2?

It's just being rolled out step-by-step, and as I wrote, we're still on the VideoFX waiting list.

You want to be on the list too? Simply go to this page, log in with your Google account and register. If needed, use a VPN with a US IP address.

With a bit of luck, you'll soon receive an invitation, fingers crossed!

2) PIKA 2.0 is here, with an ingenious feature

The second video update comes from PIKA and as promised you can try it out right here on the site.

We weren't big fans of PIKA before because it didn't make realistic videos. That has now changed.

The videos look great, but we think one feature is particularly cool: Screen Ingredients

With this new feature, you can add "ingredients" to your videos in the form of photos. This can be, for example

• Person who should appear in the video

• Clothes this person should wear

• Environment in which the person should be

You then combine these ingredients with a normal text prompt that describes what should happen in the video.

Boom! PIKA creates really good clips from this, here is the result of the three ingredients + prompt from above:

The applications here are plenty. You can make a promotional video for a product, let yourself fly, or just have fun by combining photos and bringing them to life.

Take a look at this 60-second video to get an idea. It's really cool!

The PIKA effects are also a fun feature. With just a few clicks, you can add cool effects to your videos that will definitely attract attention. Ideal for social media, ads etc.

As said, you can try out everything here immediately - the only downside: if you want to use all PIKA 2.0 features, you will need to upgrade for just about $30/month.

🎆 Google is also stepping up its game with AI images - try it out now for free

This week is a big OpenAI and Google update race. Since we've already talked about Google above, we'll continue on the same track.

Google also has a few exciting upgrades in store when it comes to images. The improved AI image model Imagen 3 delivers really cool results.

Best of all, you can try it out for free here on ImageFX. If you are in the EU, you'll need a VPN with a US IP address and a Google account.

What we like about ImageFX Editor: It analyzes your prompts and highlights properties, as you can see in red above.

When you click on the highlights, it gives you suggestions for possible variations, and below there are ideas for suitable styles.

This makes it super quick to create lots of good images and variations. As already mentioned, this is currently free of charge.

Another mini tool from Google, which we like a alot: Whisk

With Whisk, you can generate creative images from several elements with a neutral background.

These can either be reference images or text prompts, on the basis of which Whisk then generates new images.

You simply enter a reference image as the subject. This can be a selfie or, as in our case, a graphic of our mascot Nauti.

Then you define a scene and a style. How about Nauti as a cuddly toy?

Whisk then generates images from this, which you can optimize further. You can also add multiple themes, scenes, styles and let the AI get creative.

We really enjoy it, it's free, and you can try it out here.

💻 New ChatGPT Projects make the tool 3x more useful

We also have to mention a new feature of ChatGPT, which was released as part of the OpenAI Shipmas.

You may have already seen it in the menu of ChatGPT, there is a new entry called Projects.

Projects work in a similar way to those of our competitor, Claude, but we are super happy that ChatGPT has followed suit here.

In principle, they are like better GPTs that make the tool 3x more useful from our point of view, especially for longer-term and more extensive work on certain topics.

As with GPTs, you can add system instructions to projects that describe exactly who or what ChatGPT should be in the project and how it should respond.

You can also add files in the form of documents, code, images, etc. to create a knowledge base per project.

All the same as GPTs, but here are two reasons why we think Projects are so much better:

1) You can see your past chats in the project

Especially those who work a lot with individual GPTs will love it! Finally, all your chats per project are in one place, and you can access them quickly.

2) Model selection for each new chat

With the GPTs, you cannot specify whether GPT-4o or o1 is used. With a project, you can change the model per chat depending on the task.

If you need tools like Canvas etc., use GPT-4o. If you need advanced thinking and complex problem-solving, choose the clever o1 model.

All the models always have access to the instructions and knowledge.

Important to know: If you work with documents and instructions in your projects, you can only select the 4o model. However, this will hopefully change soon, as the o1 has now also been given tool access.

We love it!

P.S. For all those who use ChatGPT free of charge, there is also an important update: the web search is now also available in the free version!

We made it! But no need to be sad. The AInauts will be back soon, with new stuff for you.

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.