AI Insights
Posts
🤯 Roblox Drops AI That Builds in 3D

🤯 Roblox Drops AI That Builds in 3D

PLUS: How to Create Avatar Videos with AI

AI Insights
March 18, 2025

In partnership with

Sign up | Sponsor

Welcome, AI Enthusiasts!

Today’s AI Insights:

Roblox's new AI creates 3D objects
Mistral AI unveils small 3.1
ElevenLabs launches Scribe
Swiss startup Unique secured $30M
How to create avatar videos with AI
AI tool of the day
Latest AI and tech news
Top AI tools

^{Read Time: 3 mintue}

LATEST AI NEWS

ROBLOX

🤯 Roblox’s New AI Creates 3D Objects

Image Source: Roblox

Roblox has launched Cube, its first-ever 3D generative AI model to help creators build 3D objects using simple prompts. The company is also releasing an open-source version, making it accessible for developers beyond its platform.

The Details:

Mesh Generation (Beta): Creators can now type prompts like “an orange racing car with black stripes” to generate detailed 3D meshes instantly in Roblox Studio.
Open Source Access: Developers outside Roblox can customize, create plug-ins, or train Cube with their own datasets.
More AI Tools Coming: Roblox is also rolling out text generation, text-to-speech, and speech-to-text tools to enrich in-game interactivity with NPCs and player voice commands.
Next Step: 4D Creation: Roblox envisions a future where scenes are not only created but also react and interact — combining objects, environments, and people dynamically.

Why it matters:

Cube aims to make game development faster and more accessible, especially for indie creators. With AI, Roblox is redefining what’s possible in game design — though concerns around AI’s impact on creative jobs continue to grow across the industry.

👉 Check out Cube in action at GDC this week.

🤝 TOGETHER WITH FINANCE BUZZ

Pay No Interest Until Nearly 2027 AND Earn 5% Cash Back

Use a 0% intro APR card to pay off debt.
Transfer your balance and avoid interest charges.
This top card offers 0% APR into 2027 + 5% cash back!

Learn How To Apply Now

MISTRAL

🤖 Mistral AI Unveils Small 3.1 — A Lightweight Powerhouse

Image Source: Mistral AI

Mistral AI has released Small 3.1, a lightweight, open-source AI model that punches well above its size. With just 24 billion parameters, it rivals models from OpenAI and Google — but runs on modest hardware like an RTX 4090 GPU or even a MacBook with 32GB RAM.

This launch marks another big move in Mistral’s mission: making powerful AI models more efficient, accessible, and open for everyone.

The Details:

Small 3.1 processes text and images at 150 tokens per second and handles 128,000-token contexts, making it perfect for real-time applications.
Handles both text and images with ease — ideal for building interactive, smart applications.
Available now via Hugging Face, Google Vertex AI, and coming soon to Azure AI Foundry and Nvidia NIM.
Adds to Mistral’s growing ecosystem, including Mistral Large, Codestral (code generation), and Pixtral (multimodal).

Why it matters:

Mistral is proving that AI doesn’t have to be gigantic to be powerful. Its efficient design offers an alternative to scale-hungry models and opens up new possibilities for AI on smaller, everyday devices. More importantly, its open-source philosophy is enabling faster innovation — showing how collaboration can outpace size and funding.

AI TOOL OF THE DAY

📸 HeadshotPro: Generate Pro Headshots in Seconds

HeadshotPro is a trending AI tool in 2025 that makes pro headshots super easy!

Upload a few selfies, pick a style, and get tons of awesome photos fast—no photoshoot needed. It’s great for LinkedIn, work profiles, or just looking sharp.

👉 Snap and shine—try HeadshotPro today! ✨

AI & TECH NEWS

ElevenLabs launches Scribe, its first stand-alone speech-to-text model, supporting 99+ languages with top accuracy in 25, including English and Spanish. Outperforming Google Gemini 2.0 and Whisper, Scribe features speaker diarization and auto-tagging. Pricing starts at $0.40/hour, with a real-time version coming soon.

Hume AI launches Octave, a text-to-speech model that delivers lifelike, emotionally rich voices for audiobooks, podcasts, and games. Powered by an LLM, it adjusts tone and cadence based on context. Supports English & Spanish, with more languages coming. Available via subscription API at competitive pricing.

Swiss startup Unique secured $30M in Series A funding from DN Capital and CommerzVentures to scale agentic AI for banking, insurance, and private equity. The AI agents automate research, compliance, and KYC, with tools for investment analysis and due diligence.

Nvidia CEO Jensen Huang downplayed fears that DeepSeek’s R1 model would hurt sales, saying reasoning AI needs 100x more compute. Nvidia posted record revenue of $39.3B, with data center sales hitting $115B in 2024. Huang calls demand for the Blackwell chip “extraordinary.”

Salesforce customers can now tap into Google’s Gemini AI models, enhancing AI agents and real-time data analytics. With the agentic AI market set to hit $2T, this collaboration positions both companies at the forefront of AI-driven business transformation.

AI TUTORIAL

🎥 How to Create Avatar Videos with HeyGen

HeyGen is an AI-driven platform that crafts realistic video content using customizable avatars and natural-sounding voiceovers from simple text inputs, ideal for presentations, tutorials, or marketing clips without filming.