Grok 2.0 is wild 👀

Plus Gen-3 Turbo & Midjourney's new web editor

Aug 19, 2024

Total reading time around 5 minutes.

Welcome to Visually AI!

It’s been another busy week in generative AI, and personally - I’m writing this on the road back from moving my sons into university apartments several hours from home…

🔮AI News This Week

Grok 2.0 wins 𝕏

I’ve been using the new Grok 2.0 and it’s noticeably improved! I asked Grok 2.0 for a summary, below:

Grok 2.0, released by xAI in August 2024, represents a significant upgrade in AI capabilities, focusing on enhanced conversational skills, coding assistance, and unrestricted image generation through the Flux.1 model.

Enhanced Conversational Abilities: Improved context understanding and nuanced responses for both casual and professional communication.
Coding and Problem-Solving: Superior performance in coding tasks and complex problem-solving, outperforming some leading AI models.
Uncensored Image Generation: Utilizes the Flux.1 model for creating images with minimal restrictions, sparking discussions on AI ethics and misuse.

I don’t have data on the ‘superior performance’ but I’ve been using it to research topics posted on 𝕏 and the Flux image generation, of course!

Grok 2.0 is only available to 𝕏 Premium users for now.

Images are definitely on the ‘almost NSFW’ scale, but it still seems to be somewhat censored:

Example of Flux image generated by Grok 2.0 on 𝕏

But most of us were surprised to see Black Forest Labs’ Flux.1 model combined with Grok! We had been hearing rumors of Elon prepared to partner with Midjourney for image generation, but that seemed unlikely.

It’s great to be able to generate images with the quality and details produced by FLUX and then combining with video models, such as Gen-3 or Luma Dream Machine:

How long will uncensored-ish Grok last?

I’ve had some fun generating cool images with Grok 2.0, but I feel like this won’t last in the same form for very long.

I mean, look at some of the images in the last few days:

Screenshots of images generated by Grok 2.0 on 𝕏

Runway Gen-3 Turbo

Runway launched a faster version of their Gen-3 image to video model.

Now you can generate videos from images in less than a minute. Here’s an example with no editing:

Midjourney has a new web editor

Midjourney's new web editor introduces a unified interface for AI image editing, combining features like inpainting, outpainting, and zooming into a single, more intuitive toolset, aimed at enhancing user creativity and workflow efficiency.

This update also expands access to Midjourney's platform, allowing users who have created at least 10 images to utilize the web version for both creation and editing, signaling a move towards a more accessible web-based platform.

📸AI Snapshots

I started posting FLUX LoRA Resource threads on 𝕏 with a lot of helpful links if you’re interested in trying a finetuned version of Flux.1 or training your own. (Mega Thread 1, Mega Thread 2)

Sync.labs announced its newest lip sync model, lipsync-1.7.1-beta. It’s more than 50% faster than the previous version, more accurate and the smoothest model yet. It’s zero shot and doesn’t require training for real people or AI-generated characters.

Gemelo AI just added a feature called Twin Creator where you can upload your 5 second Kling or Runway Gen 3 video based on your Flux image and make it speak with any voice + give it a lip-sync.. for free!

Hedra launched the latest version Character-1.5 model with sharper video quality, natural blinking and head movement, wider tolerance for uploaded images with farther away objects, and a new stylization feature.

This is an example with an image I generated on FLUX Realism LoRA and ElevenLabs sound:

Google upgraded its AI image generator, Imagen 3, with improvements in image quality, realism, and editing features. It is available in ImageFX and other Google products and devices.

I got early access and it has definitely improved:

Javi Lopez, the founder of Magnific AI, has a secret image model codenamed “Mystic,” and I got a chance to try it. “Mystic” will be available on Freepik soon, and it is capable of generating high-quality photorealistic images, like this:

💻Visually AI on YouTube

Check out my new channel trailer!

The Visually AI channel includes:

IntelliVerse Podcast
AI Tutorials
Collaborations
AI Films & Entertainment
Additional AI News & Topics

🛠️ This Week’s AI Tools

Fal Flux Realism LoRA: Generate ultra-realistic images with a fine-tuned version of Black Forest Lab’s Flux.1 model. (link)

Pinokio 1-Click ComfyUI Launcher: Cocktail Peanut wrote a 1-click launcher for the new ComfyUI front-end architecture, optimized for FLUX through the Pinokio AI browser. (link)

RunwayML: There are several helpful tools available on Runway’s app, including image upscaling, video background removal, video inpainting, and blurring faces in videos. (link)

Descript: I have been using Descript to edit YouTube videos, quick tutorials, and my podcast videos & audio. I love how easy it is to edit by video timeline or transcript. (link)

Riverside FM: I’ve been using Riverside FM to record podcasts and Under the Hood episodes for YouTube. You can record up to 4k video and edit videos by updating the transcripts. It makes great clips, too. (link)

Web Designer 4000: This free Glif designs website UI images with Claude Sonnet 3.5 and FLUX Dev. You can see a few of mine below. (link)

a minimalistic AI art portfolio website for my art, called Visually AI in different colors — Prompt: A minimalistic AI art portfolio website for my art, called Visually AI

🎧IntelliVerse Podcast

🎙️IntelliVerse is now on Apple Podcasts, Spotify, and PodBean!

Listen to my fascinating discussion with Victor Perez, CEO of Krea AI on Episode 3 of IntelliVerse. (The reality TV atmosphere of his experience in a founder incubator called HF0…)

Intelliverse #3 | Inside KREA AI w. CEO Victor Perez

🖼️ Image Prompts

Prompt: A cinematic still of a man running through a narrow alleyway at night, close-up shot, cool blue and green lighting with high contrast, urgent and thrilling escape scene, shot on an Arri Alexa, 8K.

A cinematic still of a man running through a narrow alleyway at night, close-up shot, cool blue and green lighting with high contrast, urgent and thrilling escape scene, shot on an Arri Alexa, 8K. — Generated in Midjourney

Prompt: A serene woman in a flowing white dress stands barefoot by a glass-like lake, bathed in silvery moonlight. Iridescent butterflies and glowing fireflies dance around her, casting shimmering reflections. Hyperrealistic details capture every dewdrop, leaf vein, and subtle emotion in her tranquil expression.

🎞️ Video Prompt

Prompt: A volcano erupting with a flow of glitchy lava, each pixel a tiny window into different molten landscapes. The eruption cloud forms a giant, flickering digital display in the sky.

Thanks for reading, and have a creative week!