New AI Tools: Free Agent, Video Editing, and 3D Models

The AI landscape is evolving at an astonishing pace, with new tools enhancing creativity and productivity across various domains. From groundbreaking video editing software to advanced 3D modeling platforms, these innovations are reshaping how creators and developers bring their ideas to life. Let's explore some of the most exciting new AI tools that are making waves in the digital world today.

GLM 4.5: A Revolutionary Open Weight Model

The AI world has a new powerhouse model that's turning heads—GLM 4.5. Unlike many incremental improvements in language models, GLM 4.5 delivers genuinely impressive capabilities as an open weight model, meaning you can download and run it on your own hardware or cloud infrastructure.

Performance-wise, GLM 4.5 stands toe-to-toe with industry giants like GPT-4 and Claude 4 Opus, even outperforming Claude on agentic reasoning and coding benchmarks. But what truly sets it apart is its practical application and accessibility—it's completely free to use at Z.ai.

Creating Professional Slide Decks From Scratch

One of GLM 4.5's most impressive party tricks is its ability to create comprehensive, visually appealing slide decks from simple prompts. When tested with a request to create a slide deck about the satirical "Birds Aren't Real" movement, the results were remarkable:

1. The model conducted thorough research, visiting Wikipedia, the New York Times, and other credible sources.
2. It searched for and incorporated relevant images, including pictures of the movement's founder.
3. It generated professionally designed slides with proper formatting, headings, and visual elements.
4. The final result included six well-designed, cohesive slides that far outperformed what ChatGPT could produce.

Unlike other AI slide generators that might create bland, text-heavy slides, GLM 4.5 produces presentations with proper visual hierarchy, image placement, and professional aesthetics.

One-Shot Code Generation That Actually Works

GLM 4.5's coding capabilities are equally impressive. When challenged to create a "Vampire Survivors" clone in JavaScript playable in a browser, it delivered functional code in a single attempt. The resulting game featured:

• A player character that automatically fired at approaching enemies
• Enemy swarms moving toward the player
• XP collection mechanics
• Upgradable weapons, including throw knives and garlic
• Multiple firing directions as you level up

All this required just one prompt—no additional guidance or iterations. The model is exceptionally fast and delivers complex, working code effortlessly, without requiring paid features.

Revolutionary AI Video Editing Tools

The advancements in AI tools extend beyond text generation, especially in video editing.

Runway's ALF: Targeted Video Modifications

Runway has introduced ALF, a groundbreaking feature that allows for precise, targeted edits to video content based on natural language instructions. This tool enables users to modify specific elements within videos while preserving the original content.

When tested with various clips, ALF demonstrated impressive capabilities:

• Transforming environment contexts, such as placing a jet fighter from Top Gun Maverick into space
• Replacing objects within scenes, like swapping babies for a sandwich in The Flash
• Creating new camera angles from existing footage
• Combining screenshot animations with 3D camera movements for product demos

ALF excels at changing environments and swapping individual elements, though more complex scene transformations can sometimes yield unpredictable results.

Luma Labs' Modify With Instructions

Luma Labs offers a similar feature called "Modify with Instructions," allowing for natural language editing of video content. The tool showed promise in testing, with capabilities that included:

• Successfully placing the Top Gun jet in space with an alien spaceship chasing it
• Preserving original audio while modifying visuals
• Providing adjustable "strength" settings that determine how closely the output adheres to the original video

With this strength slider, users can easily control the extent of modifications—creating entirely new videos at maximum strength, while maintaining more of the original footage at lower settings.

Google's V0 Model and Emergent Text-Following Behavior

Google's innovative V0 video generation model has revealed fascinating emergent behavior—it can follow text instructions placed directly on input images. This unexpected capability allows users to create an image with written instructions, submit it to the V0 model, and watch as the generated video follows these instructions step by step.

Examples include:

• Instructions for a dune buggy racing toward the camera, followed by a helicopter dropping into the scene
• A man jumping on a giant turtle that then heads to the sea
• A man jumping on a crocodile that opens its mouth

This technique provides a novel way to control video generation with precise timing and sequencing of events.

MidJourney's Start/End Frame Animation

MidJourney has introduced a new feature allowing users to specify both starting and ending frames for video generation, opening up exciting possibilities:

• Create seamless transitions between two different scenes
• Generate looping videos using the same image as both start and end frames

This innovative approach grants users more control over animation endpoints and enables creative transitions between scenes, though complex transformations (like morphing between radically different subjects) remain challenging.

Idogram Character: One-Shot Face Swapping

Idogram has released a remarkable feature that allows users to place any face on any image using just a single reference photo. This represents a significant advancement over previous approaches that required 20 or more training images.

The system works impressively well by allowing you to:

• Upload a single face photo
• Select from templates or upload your own target image
• Apply your face to the target with minimal effort

The technology maintains facial expressions and integrates naturally with the target image's lighting and composition.

Meshy 5: Advanced 3D Model Generation

Meshy has unveiled version 5 of their 3D model generation platform, boasting significant improvements to model quality and detail. The system can:

• Generate detailed 3D models from text prompts, such as a pizza with multiple toppings
• Convert 2D images into 3D models with remarkable accuracy
• Maintain complex textures and details

While not fully production-ready for all applications, the quality of outputs marks a major step forward for AI-generated 3D assets.

Additional AI Advancements

The landscape continues to evolve with several notable developments:

Adobe Photoshop's New AI Features

Adobe has added several powerful AI features to Photoshop Beta, including:

• Generative Upscale: Dramatically increase image resolution while preserving and enhancing detail.
• Harmonize: Automatically match lighting, color, and tone when compositing elements from different images.
• Improved Remove Tool: More accurate object removal and better background regeneration.

The Harmonize feature stands out, seamlessly blending subjects from one lighting environment into entirely different scenes.

Other Notable Developments

• Google AI Mode in the UK: Google's AI-powered search with multimodal capabilities has expanded to the UK.
• Notebook LM Video Overviews: Google's Notebook LM now generates visual slideshows to accompany AI-generated audio.
• Microsoft Edge Co-Pilot Mode: Edge browser incorporates agentic features akin to specialized AI browsers.
• Fable - "Netflix of AI": An Amazon-backed company developing technology to generate full-length cartoon episodes based on prompts.
• Higsfield's Halo AI Video: A powerful Halo AI video generation tool available for free (at least temporarily).
• Unitree R1 Robot: A humanoid robot priced at $5,900 capable of impressive movements like cartwheels and handstands.

As the pace of AI advancement continues to accelerate, each week brings tools that transform what's possible for creators, developers, and everyday users alike.

Conclusion

Embrace the future of creativity and innovation by exploring these groundbreaking AI tools today. Whether you're looking to enhance your video editing skills or generate stunning 3D models, there's no better time to dive in. Visit Z.ai to test GLM 4.5 for free, and stay ahead of the curve with the latest advancements. Don't miss out on the opportunity to revolutionize your projects—start experimenting now!