Runway's Aleph: Text to VFX

In partnership with

Welcome to VP Land! Runway has unveiled Aleph, an in-context AI video model that combines object manipulation, scene generation, and style modification into a single, unified system. This could be a massive bridge for blending the power of generative text-based editing with existing video footage.

Last week, we asked for your thoughts on the proposed AI copyright bill requiring explicit consent for training data. Responses were mixed, but most of you said the real focus should be on the outputs, not the inputs. Check out today's poll below.

In today's edition:

Runway launches Aleph AI video model
Alibaba upgrades open-source Wan 2.2 AI video suite
Tencent unveils text-to-3D world generator HunyuanWorld 1.0
Creator details multi-tool AI workflow for viral video

Runway's Aleph: Type Out What You Want to Change

Runway has launched Aleph, an in-context AI video model that consolidates object manipulation, scene generation, and style modification into one unified system. The platform enables creators to add, remove, or transform objects within existing video clips, as well as generate new camera angles from existing footage through simple text commands.

Aleph addresses the challenge of using multiple specialized tools by offering what Runway calls "multi-task visual generation," allowing creators to perform object removal, style changes, and scene extensions within a single contextual workflow.
The system demonstrates advanced 3D video understanding that synthesizes different camera angles and virtual cinematography while maintaining temporal coherence when inserting or removing elements across video frames.

While Aleph is rolling out to users this week, a few demo cases posted online include:

Video relighting
Wardrobe and styling modifications
Instantaneous inpainting
Video transformations like removing backgrounds, changing the weather, and more

SPONSOR MESSAGE

Create How-to Videos in Seconds with AI

Stop wasting time on repetitive explanations. Guidde’s AI creates stunning video guides in seconds—11x faster.

Turn boring docs into visual masterpieces
Save hours with AI-powered automation
Share or embed your guide anywhere

How it works: Click capture on the browser extension, and Guidde auto-generates step-by-step video guides with visuals, voiceover, and a call to action.

Best part? It’s 100% free.

👉 Download the extension now

Alibaba's WAN 2.2: Open Source AI Video Gets Cinematic Quality Boost

Alibaba's Wan team has released Wan 2.2, a major upgrade to their open-source AI video generation suite, introducing three new models with an advanced Mixture-of-Experts architecture. The release brings significant improvements in video quality, resolution support, and creative control for filmmakers and content creators.

Mixture-of-Experts architecture divides the denoising process across specialized expert models handling different noise levels, dramatically increasing model capacity while maintaining the same computational cost as previous versions.
The new model lineup includes three variants: Text-to-Video models (A14B and 14B), Image-to-Video (A14B), and Text-Image-to-Video (5B), each supporting native 1080p output and seamless integration with ComfyUI and Diffusers workflows.
Training data expanded by +65.6% more images and +83.2% more videos compared to Wan 2.1, with curated aesthetic datasets featuring detailed labels for lighting, composition, contrast, and color tone.
The 5B Text-Image-to-Video model runs efficiently on consumer GPUs like the RTX 4090, achieving 720p at 24fps while supporting both text prompts and image guidance for precise creative control.
All models come with multi-GPU inference code, full Apache 2.0 licensing for commercial use, and immediate availability of model checkpoints for production deployment.

Tencent Creates Text-to-3D World Generator

Tencent Hunyuan launched HunyuanWorld 1.0, a framework that generates immersive, explorable 3D worlds from simple text descriptions or images. The announcement on X showcases how creators can now build complete virtual environments without traditional 3D modeling skills.

We're thrilled to release & open-source Hunyuan3D World Model 1.0! This model enables you to generate immersive, explorable, and interactive 3D worlds from just a sentence or an image.

It's the industry's first open-source 3D world generation model, compatible with CG pipelines
— #Hunyuan (#@TencentHunyuan)
2:01 AM • Jul 27, 2025

Panoramic world generation serves as the foundation, creating 360° images that act as proxies for full 3D scene reconstruction across diverse artistic styles and environments.
The system uses semantic layering to automatically decompose scenes into sky, background, and interactive object layers, enabling creators to manipulate individual elements within generated worlds.
Direct mesh export makes the technology compatible with existing game engines like Unity and Unreal Engine, plus VR platforms including Apple Vision Pro and Meta Quest.
Performance benchmarks show HunyuanWorld 1.0 outperforms existing methods like WonderJourney and DimensionX across visual quality metrics while maintaining strong alignment with input prompts.
Applications span virtual production workflows, from creating VR environments and game assets to enabling physical simulations with collision detection and fluid dynamics.

AI Creator Breaks Down Viral Super Mario Transformation

Creator enigmatic_e shared a detailed breakdown of how they created their viral Mario transformation video, revealing a sophisticated multi-tool AI workflow that's capturing attention across creative communities.

As silly as this concept is, I’m super proud of it 😅
— #enigmatic_e (#@8bit_e)
5:10 PM • Jul 25, 2025

Step 1: Sora served as the primary reference generator, creating both the initial Mario outfit design and later producing a "beefy" muscular version for the transformation sequence.
Step 2: They used ComfyUI as their main workspace, running multiple AI models including FLUX Kontext for background replacement and Wan 2.1 VACE for video-to-video transformation.
Step 3: Face preservation became critical when the AI changed the character's appearance too much, requiring traditional rotoscoping techniques to create masks that kept the creator's face intact while blending AI-generated elements.
Step 4: Openpose Editor by toyxyz helped resize body proportions when the initial muscular transformation wasn't dramatic enough, showing how creators combine specialized tools for specific adjustments.
Step 5: The final workflow included Pikaframes for smooth transitions and After Effects for traditional post-production polish.

BTS look at the upscaling and outpainting work done to The Wizard of Oz to prepare it for screening at the Las Vegas Sphere as a new immersive 16K experience.

Stories, projects, and links that caught our attention from around the web:

🌍 Meta is aiming to create a photorealistic digital replica of the real world for Horizon, using photoreal Gaussian splatting technology.

🎲 Board game crowdfunding platform Gamefound has acquired OG crowdfunding website Indiegogo.

🛠️ Google's experimental new platform, Opal, allows developers to build AI-powered mini-apps using natural language instructions.

🕶️ Chris Samra has launched Waves, camera glasses specifically designed for creators seeking to record and livestream content discreetly. What could go wrong?