AI Video Wars include Veo 3.1, Sora 2, Ray3, Kling 2.5 + Wan 2.5
House of David Season 2 is here!
Welcome back to Visually AI!
Today’s reading time is 5 minutes.
It’s been longer than I planned since the last edition, but there is a lot of information in the AI video category. Veo 3.1 launched today, as I prepared to publish this article!
In today’s edition:
Veo 3.1 brings richer audio and object-level editing to Google Flow
Sora 2 is here with Cameo self-insertion and collaborative Remix features
Ray3 brings world-first reasoning and HDR to video generation
Kling 2.5 Turbo delivers faster, cheaper, more consistent results
WAN 2.5 revolutionizes talking head creation with perfect audio sync
House of David Season 2 Trailer
HeyGen Agent, Hailuo Agent, Topaz Astra, and Lovable Cloud updates
Image & Video Prompts
People often ask me which video model is my favorite or the ‘winner’. I have several that I use frequently but those change, depending on improved versions or new models that I have tested or specific use cases.
Right now, I’m using Kling 2.5, Wan 2.5, Veo 3 / Veo 3.1, Seedance Pro, and Sora 2 for a lot of regular content and testing. I don’t bother choosing a winner because that could change week-by-week.
Here’s a look at the latest video models and updates:
Veo 3.1: Improved over Veo 3
Google’s newest video model, Veo 3.1 is here, and it’s better than the previous Veo 3 model (not a huge difference yet…)
What’s new and/or improved:
• Narrative Control - Use “Ingredients to Video” with multiple reference images to control characters and style.
• Seamless Transitions - “Frames to Video” lets you define the start and end of a shot for perfect transitions.
• Richer Audio - Native audio generation is now part of all features, including dialogue and sound effects.
• Object-Level Editing - Add or remove elements from your scene with natural-looking results.
Available now in Google Flow, the Gemini app, and through the Gemini and Vertex AI APIs. Also available on several platforms including Adobe Firefly Boards, Freepik, Replicate, and more.
Sora 2: The “world simulator” is here
OpenAI just dropped Sora 2, and it’s being called the “GPT-3.5 moment” for video. Welcome to a new way of thinking about AI video, with a focus on physics, audio, and social creation.
Capabilities:
• Synchronized Audio-Video - Generate video and audio (dialogue, SFX, music) together with your video.
• Realistic Physics - Sora 2 understands how objects and people move in the real world.
• Cameo Feature - On the Sora iOS app, you can quickly record yourself on your phone for a few seconds to put yourself or public Cameos into any scene.
• Remix and Collaborate - Build on other users’ creations, changing the style or setting while keeping the core action.
So far, it has an interesting social media impact with a new social feed in the Sora app - you can like and comment on posts, follow creators, and choose to leave videos as drafts or post to your profile.
Available now via an invite-only iOS app and web platform. A free tier is available, with a “Pro” version for ChatGPT Pro subscribers.
Here’s an example with my own Cameo using this prompt:
hbcoop walking down a 2D animated world in a city with 2D characters walking in the scene, but she is photorealistic
Ray3: AI that actually thinks before creating
Luma just dropped Ray3 and it’s impressive. This is the world’s first reasoning video model where it thinks through your prompt, judges its work, and tries again until it gets it right.
What’s inside:
Chain of Thought reasoning - no more random generations
Native 16-bit HDR - export EXR frames for pro workflows
Draft Mode - 5x faster, 5x cheaper for rapid iteration
Visual annotations - draw on images to control motion
Draft Mode lets you sketch ideas in seconds, then use Hi-Fi to master the best shots into 4K HDR.
This is huge because it’s an AI video that collaborates instead of just endless generations trying to get something right. The reasoning system means fewer wasted credits and more intentional results.
Available now in Dream Machine, Adobe Firefly, and other platforms.
Here’s a mock trailer I created using Ray3:
Kling 2.5 Turbo: Better, faster, cheaper
Kling’s new 2.5 Turbo is hitting all the right notes - 30% cheaper than 2.1 with major quality upgrades.
Kling has been one of my go-to video models for over 1 year.
Key improvements include:
Better physics - realistic cloth, fluid motion, lighting
Improved prompt following - your vision translates better
Enhanced stability - more consistent generations
Faster processing - less waiting around
The pricing win: 5-second 1080p videos cost significantly less, making regular content creation actually sustainable compared to Kling 2.0 and 2.1.
WAN 2.5: Generate with synced audio
Alibaba’s WAN 2.5 emerged as an alternative to Veo 3 for videos generated with audio-visual sync.
Features include:
Native audio generation - voice, SFX, music all synced
1080p at 24fps - up to 10 seconds duration
Great lip-sync - looks natural
Multi-modal input - text-to-video and image-to-video
Text writing - incredibly accurate text rendering in video
Available on Higgsfield, Freepik, Krea, ComfyUI.
🎥 House of David Season 2
I’m so proud to have worked on Wonder Project’s House of David Season 2 production on location in Greece with an amazing VFX AI crew including my friends: Ludovic Carli, Pierrick Chevallier, and Julien Cocquerel!
Some of you may know that I worked on the post-production VFX AI team for Season 1, which aired earlier this year. Season 2 was different because we worked during production at the studio in Greece. For the new season, we generated many LED wall assets used as immersive sets for certain scenes shot with actors and crews at the studio.
Check out the Season 2 trailer for the show which premiered on October 5 for Amazon Prime Video (US only) below - check out some of the battlefield background scenes for our work that was used on the LED walls during live shots:
🚀 My Recent Top AI Tools Picks
HeyGen Agent (Public Beta): HeyGen’s new Video Agent transforms single prompts into complete professional videos with script, visuals, avatar, voice, pacing, and captions.
Hailuo Agent: MiniMax’s creative sidekick turns ideas into visual content on an infinite canvas. The agent selects optimal models, plans workflows, and combines images, videos, and audio in one integrated platform. Currently offering a free week trial for new users.
Topaz Astra: Part of Topaz Labs’ new Studio ecosystem, Astra offers precise and creative modes for enhancing videos to pristine 4K quality. The model is designed specifically for iterative AI filmmaking processes and integrates with the broader Topaz enhancement suite.
Lovable Cloud & AI: Launched September 29th, this platform now enables full-stack, AI-native app development through simple prompting. The integrated backend handles databases, authentication, and file storage, while Lovable AI (powered by Google Gemini) adds AI functionality without API key management.
🎨 Explore my Portfolio
I’ve had some amazing opportunities to work on a variety of projects recently, and I don’t share everything in this newsletter or online. And you know I love testing tools to share.
Take a look at my portfolio to get a quick glimpse of my work:
🖼️ Image Prompts
Prompt: A group of travelers walking together through a marketplace, shadows long against the walls.
Prompt: A forest clearing with the silhouette of a deer, the backdrop a dazzling display of autumn foliage in oranges and reds giving way to the cool, misty blues of a distant mountain range.
🎥 Video Prompts
Prompt: A skateboarder launches off a half-pipe, performs a 360 kickflip midair, lands smoothly, and raises their arms to the cheering crowd.
Kling 2.5 Turbo
Prompt: First-person POV racing on a motorbike through a mountain road, leaning into sharp turns, wind rushing past, headlights streaking across the dark tunnel ahead.
Luma Labs Ray 3
Thank you for reading.
I hope you have a creative week!
Heather Cooper
A dad from Portugal, amazing post congrats, and the serie House of David love it... The S01 and now i am on S02E03 great on did