StyleDrop and SnapFusion

Hey there,

Apple’s next big thing is finally here! Mixed-Reality headsets are expected to be launched today.

We're all wondering, will it shake things up like the iPhone did in 2007?

Meanwhile, Meta grabbed the opportunity to unveil Quest 3 headset last week, right before Apple’s debut.

Moving on to today’s update:

📰 News: New optimized AI model can run on your phone!

🛠️ Tools: AI-narrated tours and Clone your audio in other languages using same voice

🧠 Learn: Google launched 9 free courses on Generative AI!

🧠 Misc: Writes are loosing their jobs over AI and read how digital doubles could work in hollywood.

Reading time is 5 minutes. Let’s dive in 🤿

SnapFusion: AI on your phone

Until a month ago, you’d need a powerful PC with beefed-up GPUs to run an AI model locally on your computer (and it was still slow!)

Now you could run a text-to-image AI model on your phone, generating a custom image in less than 2 seconds!

Snap released a new AI model called SnapFusion that optimizes the de-noising process, requiring only 8 steps instead of 50 in the Stable Diffusion v1.5 while achieving better FID and CLIP scores!

Why it matters: The capability to run AI models locally on your devices is a game changer for privacy and security of your data.

For instance, your sensitive data like your images and voice never leaves your device. Plus, you get the benefits of real time inference, which could amp up AR/VR experiences on your phone.

Style Drop: Transfer any style

StyleDrop by Google can help you generate images from text in any style using a single reference image!

The technique is called Style Transfer and has been around for a while. But StyleDrop drastically cuts down the number of reference images to just one!

The AI models can extract color scheme, shading, design and even format the output images to fit the style.

Why it matters: Any creative task on images usually involves generating multiple styles and variations to see what works best and fits the overall scheme of the project.

StyleDrop could generate those variations within seconds, helping all creatives, from graphic designers to video editors, to iterate faster and optimize their workflow.

New Tools
Leverage AI to grow

• Narrated Tours: AI generated custom audio tour guide for your city tours (link)

• Seodity: All-in-one SEO tool with better domain analysis, AI content creation, easy teamwork, and helpful automation.(link)

• Dataspot: A research assistant that helps you to simplify complex topics, summarize content and get answers immediately. (link)

• Clonedub: Convert your audio into other languages using the same voices (link)

Learn

• FemtoGPT: Pure Rust implementation of a GPT language model (link)

• OpenAI’s Shap-e tool to create 3D shapes and images using text input (link)

• Google’s new learning path on Generative AI with 9 free courses! (link)

Misc
From the corners of the Internet

• Read: Should tech companies define regulations on AI? (link)

• Google launched AI-generated answers in search last month, but its too slow to be usable (link)

• How AI-generated video is changing the film industry (link)

• Goldman Sachs predicted 18% of jobs could be automated by AI: Writers are already loosing their jobs! (link)

• Digital doubles: AI could allow actors to work on several projects at the same time (link)

• Manage the Generative AI risk before it manages you (link)