Meet Luma's new AI video tool Ray2
MiniMax-01 launches open source foundational language model with a 4M-token context
Total reading time around 5 minutes.
Welcome to Visually AI!
Happy New Year! I hope 2025 is treating you well, so far.
I’ve been busier than usual for the past 2 months working full-time with a team on a large project that I’ll be able to share more details about… soon. 🤐
The generative AI world hasn’t slowed down, so let’s get into it!
🔮AI News this Week
Luma AI Ray2
Luma AI's Ray2 is an advanced text-to-video generative model that produces realistic visuals with natural and coherent motion, making it a powerful tool for filmmakers, marketers, and content creators.
It features a new multi-modal architecture that enhances motion fluidity, realism, and logical sequencing, significantly improving production quality and efficiency.
Features:
Text-to-video generation: Create videos from text prompts.
Enhanced motion and realism: Produces fast, coherent motion and ultra-realistic details.
Versatile input options: Supports text, image, and video inputs.
Upcoming features: Image-to-video, video-to-video, and editing capabilities on the way.
Improved production readiness: Higher success rate of usable generations for more polished results.
Ray2 is currently available to paid subscribers on Luma AI Dream Machine. Watch the official trailer here.
Here’s a few of my first results:
📸 AI Snapshots
Hailuo AI launched MiniMax-01, an open source foundational language model with a 4M-token context built on a Lightning Attention model architecture.
MiniMax-01 can also search the web. You can try it for free on Hailuo AI Chat and Hugging Face, or run it locally - GitHub.
I asked MiniMax-01 to analyze a list of effective video prompts shared by my friend, LudovicCreator on 𝕏, for important keywords, phrases, and alternative terms I could use to improve my own video prompts:
OpenAI is rolling out ChatGPT tasks that allows you to schedule recurring actions, reminders, news summaries, or other things for a future time. The new feature will be available to Plus, Pro, and Teams users, and eventually to all ChatGPT users.
I scheduled a weekly generative AI news list for every Thursday morning:
Runway introduced a 4k upscaler to Gen-3 Alpha and Turbo. It works on videos generated by Gen-3, so you can’t upload a video for upscaling. It costs around 20 credits per 4k upscale, and it depends on the length of the video (I haven’t been able to locate accurate credit information, yet).
📱 Recently On 𝕏
Tutorial: AI Video Prompting
I posted a tutorial about AI video prompting for photorealistic scenes because AI video models often struggle with several things:
Skin texture
Subtle movements
Expressions of emotion
Realistic raindrops interacting w/ environment
To get the best results, you have to do a little more prompt work. So I studied the image and tried to address each of those problems through prompting. The example below is a Midjourney image animated with Kling 1.6:
There isn’t a standard prompt structure, although it doesn’t hurt to think about building prompts using:
[Camera angle / shot type], [subject], [subject description], [setting / scene], [ action], [additional details, tone, or mood]
Each image input is unique and AI can render a different result every time, so be flexible and try different ideas, keywords, or alternatives to replace or reinforce your description of what you want to see.
🚀 My Recent Top AI Tools Picks
Hugging Face AI Agent Course: Hugging Face launched a FREE course on LLM Agents to learn how to build your own Agents and receive a certificate of completion.
Runner H: Advanced AI agent for real-world applications handling small tasks, like automations, to building websites, finding available domain names, and performing market research to create your own business. Join the waitlist.
GitHub Copilot: GitHub’s coding assistant is now open without a waitlist.
Ludus AI: New Unreal Engine ‘development companion’ generates code, creates 3D/2D assets, supports Blueprint development, and more.
Krea Video-to-Audio: Add sound to your video with one click to videos generated with top AI video models on Krea Video. Check out my demo of this feature, below:
Google AI Studio: Google Gemini can now analyze videos, along with images, docs, and text. I’ve been using it frequently on the free plan with the Gemini 2.0 Flash Thinking Experimental model.
Freepik Video Sound: Add sound to videos generated on Freepik’s AI Suite with a text prompt.
MMAudio: Upload videos to generate synchronized audio to match your content.
Bria Generative Fill API: Hugging Face demo for Bria’s simple and effective generative fill tool, lets you upload an image to erase or add elements.
Hailuo AI Audio: MiniMax launched T2A-01-HD, its latest Text-to-Audio model with improved emotional depth, versatility, and fluently multi-lingual across 17+ languages with natural accents. It’s available free for a limited time.
Sync lipsync-1.9.0-beta: Sync announced its new lipsync-1.9-beta model with zero-shot ability to generate and edit natural voices without training data and improved lipsync quality.
I cloned my own voice on Hailuo AI Audio with it’s new voice model and lip synced using Sync’s new 1.9-beta (video generated with my custom Flux LoRA image and Kling 1.6 on Krea AI):
🖼️ Image Prompts
Prompt: A young woman with curly dark hair wearing a modern denim jacket over a pastel blouse, sitting in a cozy diner with soft vintage lighting, warm tones, and a blurred background of glowing lamps
Prompt: A soaring eagle flying above a vast canyon, its wings fully extended as it glides over rugged cliffs and winding rivers below, with the warm glow of sunset highlighting the dramatic landscape
🎥 Video Prompts
Prompt: Crystal clear water flowing over smooth river rocks, sunlight creating sparkles on the surface. Low angle shot, natural movement, pristine quality
Generated on Hailuo MiniMax:
Prompt: Fog rolling through ancient stone archways in a medieval town at dawn. Tracking shot, mystical atmosphere, soft diffused lighting
Generated on Luma Ray2:
Thank you for reading.
Have a creative week!
