Hey there!

Welcome everybody! It’s Monday and we all know what that means: AI BITS is dropping a newsletter! We hope you will enjoy this article’s news and tools; we are sure they will not disappoint you😏.

Here’s what we’ll learn about today:

• Italy welcomes back ChatGPT 🍕

• Everyone can speak English with Play.ht 🗣️

• The wonders of DeepFloyd 🎨

Reading time: 4 minutes and 40 seconds

News Reel:

Highlight 1: ChatGPT is back in Italy!

A couple of newsletters ago we told you, Italy was taking a break from ChatGPT due to privacy concerns…but not anymore! 🥳

Italians can finally go back to ask AI recommended pasta recipes 😀 🍝

The Italian watchdog (Garante - The Italian Data Protection Authority) asked OpenAI to temporarily stop processing the personal information of Italian users while they checked out a possible data breach. The Garante emphasized how important it is to follow the EU's data privacy rules, but they also said they didn't want to get in the way of AI's development.

OpenAI has made some changes to make sure everything's cool now:

• They've added more info to their website about how they collect and use data to train algorithms. algorithms

• They've given EU users a new way of objecting to having their data used for training🚫.

• Plus, there's now a tool to check users' ages when signing up. 💻

The Garante said they're happy with the changes OpenAI made, but they want the company to do two more things:

• Add an age-verification system 

• Launch a campaign to inform Italians about their right to opt out of data processing.

The return of ChatGPT to Italy is a positive step towards responsible and ethical AI development.

So, to all of our Italian friends reading from Italy: What are your thoughts on ChatGPT’s comeback? What did you miss? 💥

Highlight 2: No English? No problem!

Have you ever faced language barriers with English…we’ve got news for you! Play.ht has just launched its latest machine learning model that supports multilingual synthesis and cross-language voice cloning!

😲 Fancy words but what does that mean?

Well, now you can upload a 30-minute audio of someone speaking in any language (in any specific accent you prefer) and have the software clone your voice and can generate audio of you speaking perfect English.

Imagine the immense potential! For example language learning, dubbing, localization, and more! 🌎

The most impressive thing about this model is that it does not require any sort of transcript as input for it to work. In fact, the software identifies the words in the initial audio’s voice and converts them to English, while keeping the accent and nuances of the original input’s language. 🔊

But how can we use this amazing tool? We’ve got you covered! Just follow the next steps:

• Create an account in Play.ht and click on the Voice Cloning Section

• Submit your audio speaking in your preferred language. There are two options: High Fidelity (which takes a bit longer to convert), and instant cloning (which is faster).

• Click on Create Audio back in the dashboard.

• Select your recently cloned voice from the pop-up menu.

• Finally, enter the English text in the Rich-Text editor and you will have your English-speaking cloned voice ready in a couple of seconds!

If you are looking for an example of what this powerful tool, Play.ht has released an impressive video in its web. We will leave it down below so you can have a look at it 👇🏼. Enjoy!

A.I. Tool Showcase:

DeepFloyd:

New Open-Source Model Release Alert! 📢 

Stability AI, along with their DeepFloyd multimodal AI research lab, has just announced the release of a new text-to-image🖼️ model called DeepFloyd IF.

Here are some of the features of DeepFloyd IF ⚠️:

• It can incorporate text descriptions into images with various objects appearing in different spatial relations. ⬆️ This is huge as currently, available models are really bad at this

• It has a really high degree of photorealism

• It can generate images with non-standard aspect ratios, including vertical and horizontal.⏹️

• It can modify images without fine-tuning, allowing you to change the style, patterns, and details while maintaining the basic form of the source image.

In the picture below, we created a Chinese restaurant with the name of our newsletter, and the prompt was just “A Chinese restaurant with a "AIBits" neon sign. Cyberpunk photography”! However, the DeepFloyd community is still searching what are the best ways to get the most out of this mind-blowing feature.

And the best part of everything: It is Open Source! This means anyone can use it for free.

To do so you can go to DeepFloyd’s Hugging Face space using this link. This site is used to deploy Machine Learning models so that people can use them without having to train the engines themselves. For those interested in the code, you can access the GitHub repository here! 🧑🏽‍💻

Did you know that “prompt-engineers” is now a proper job where you write and improve AI prompts to get the best output possible? 🪄

Image generation Guide

This is a short guide to creating your own images. 🎨

• Enter the prompt you would like to generate in the text box and click on “Generate”.

• Optional: You can also enter a negative prompt, which will tell the model what you would NOT want in your generated image.

• The model will display different pictures, click the one that you prefer and click “Upscale”.

• This will give you a more detailed version of the chosen image! 🖼️

A.I. Art of the Week:

See you soon!

-A.I. Bits