[AI] An Overview of New AI Tools
New segment, photorealistic faces, AI characters, AR + AI, 2D image movement, real-time data & more
It’s no surprise that the OpenAI chatbot, ChatGPT, has become the all-consuming target of adoration. It can summarize text, code, write stories, create games, turn into a virtual machine, etc. The platform has reached 1 million users 60x faster than Facebook. It seems that AI is here to stay this time around.
The AI ecosystem is growing by the day and new products are released daily. Below are just some of the available tools.
New Segment
In this new AI segment of the newsletter, I’ll be providing you with only the most interesting and actionable AI insights and tools so we can grow collectively. It may quite literally be an existential issue to stay on top of the updates. I’ve been looking into AI solutions for a while now (and have even written about it). As we move forward, I’ll be looking to integrate AI content into the context of NFTs and fashion. In short,
I’ll write about what I learn
The intention isn’t to virtue signal but to provide you with useful content
Like you, I’m still learning as we go, but if you feel left behind feel free to DM for any questions.
Without further ado, here are some of the highlights in the AI space I discovered over the past week or so:
Photorealistic Faces
One of the weaknesses of text-to-image has been creating photorealistic, non-distorted faces. Yet, you may have seen posts like these with incredibly realistic faces:
One of the keys to generating such photorealistic models is to include negative Stable Diffusion prompts. These prompts often include the terms “disfigured", “cartoon” and “blurry”.
Here’s a link to a free Stable Diffusion Colab used to generate this.
Alternatively, in case you’ve already created a great image but the faces are messed up, you can use a tool like the Applied Research Center (ARC) face restoration tool.
AI Characters
We can now chat with AI versions of anyone from Elon Musk to Plato. While you can do this manually via ChatGPT, platforms like Character AI and Inworld Studio. provide an intuitive (and social) interface to interact with these characters and even chat voice chat. In the context of crypto, this is especially useful for play-to-earn (or play-and-earn) games. Using this technology, in-game characters should have much more depth to their personality with much less effort from the creators. Moreover, training/creating own characters could become huge in social media. Perhaps, linking digital avatars like Lil Miquela, with interesting personalities will be what will drive the next iteration of social media influencers. There might even be potential for a train-to-earn model.
AR + AI
Recently, Ian Curtis from the WebAR (Web-based AR) company 8th Wall, created a captivating AR experience that utilized traditional AR tools in combination with image textures generated via AI. This reminds us of just how synergistic the AR + AI relationship is becoming:
Real-time AR: ChatGPT + WebAR
Video-to-3D asset: NeRF technology (e.g. LumaAI):
2D text-to-image to AR
While the AI characters give virtual characters social attributes, turning a video into a 3D asset will create the characters’ physical attributes. Especially the relatively lean crypto teams will benefit from the saved time and reduced need for resources. I’ve written an in-depth piece on the significance of AR + 3D for fashion brands and NFTs if you’re interested in how these technologies can be applied to fashion specifically.
Creating Movement for 2D Images
Another way you can extend the capabilities of AI characters is by making your avatars speak. This will be big for creating online content. While human influencers will likely maintain an edge, especially traditional “faceless” companies would benefit from the increased availability of digital influencers.
Movement is not restricted to avatars but can be applied to landscapes as well. While we’re waiting for text-to-video AI capabilities, we can use this GitHub code to fly into 2D images (see video below). This model was recently published by Google.
Access to real-time information
One of ChatGPT’s limitations is its inability to browse the internet and the fact that its training data only goes up until 2021. Because of this, you can’t ask topical questions like what’s the current price of Bitcoin. Accessing real-time information drastically extends the use case of the technology. To this end, there are two solutions to this that I’ve come across:
ChatGPT Assistant: An extension that allows ChatGPT to look access websites by using a simple /fetch command
Hello: A search engine combined with ChatGPT
Other Tools
Notion introduces an AI writing tool. It’s currently in private alpha and can be accessed via a waitlist. It allows users to summarize, edit and come up with ideas.
ChatGPT creates tailored weight loss plans
Wrap Up
There are still many interesting tools that I didn’t have the space to cover and will be writing about in the future. I also wanted to include my own outputs. I’ll include them in my next AI post which will focus on text-to-image generation. Here’s one for the road.
Check Out
The 27-page NFT strategy report I did for the design house Marimekko
Articles
Tools
Socials