🚨Meta launches its own AI infrastructure initiative as Zuckerberg bets that controlling energy capacity and data center buildout will prove more defensible than competing on models alone—signaling the AI race has shifted from who builds the best algorithms to who secures the physical infrastructure to run them at scale.

What if you woke up tomorrow with all your expenses covered for an entire year? No rent. No bills. What would you dream up? What would you build?

Our Dare to Dream challenge answers these questions. We believe in Creators, in Entrepreneurs, in the people who bet on their own ideas and their will to make them real.

That’s why we’re awarding $100,000 to one person who shows up as their authentic self and tells us how their dream can make a real difference in their communities.

We’ve got five runner-up prizes each worth $10,000, too. So get out your phone, hit record, and dream the dream only you can dream up.

Today is your day.

Sign Up

_{NO PURCHASE NECESSARY. VOID WHERE PROHIBITED. For full Official Rules, visit daretodream.stan.store/officialrules.}

For months, AI video creators faced the same maddening problem: you'd generate a character you loved in the first frame, only to watch them morph into a completely different person by frame 50. The face would shift. The hair would change color. The outfit would transform. Your protagonist became a shapeshifter—and not in a good way.

This inconsistency killed AI video before it could even start. You couldn't tell stories. You couldn't build brands. You couldn't create anything professional because your character wouldn't stay the same person for more than three seconds.

Then Google released two tools that changed everything: Nano Banana Pro (Gemini 2.5 Flash Image AI), designed to solve the character consistency problem, and Veo 3.1 with its start frame and end frame control. Suddenly, creators could lock in a character's identity and place them across multiple scenes, videos, and narratives without the uncanny morphing that plagued earlier models.

This isn't just an incremental improvement. It's the difference between AI video being a novelty and AI video being a production tool.

The Problem That Held AI Video Back

One of the biggest challenges in AI image generation has always been consistency. You create a character you love—only to find that the next image generated makes the face slightly different, the hairstyle altered, or the outfit unrecognizable. For storytellers, designers, and marketers, this inconsistency is a major barrier to professional use.

Think about what this meant in practice: comic creators couldn't make sequential panels. Marketers couldn't build brand mascots. Game developers couldn't generate character assets. YouTubers couldn't create recurring personas. Every frame was a gamble.

The traditional workaround was painful: generate hundreds of images, cherry-pick the ones that looked similar enough, manually edit in Photoshop to force consistency, and hope your audience didn't notice the subtle drift. It wasn't sustainable. It wasn't scalable. And it definitely wasn't professional.

How It Works: The Reference Image Revolution

Both Nano Banana Pro and Veo 3.1 solve this with the same core insight: stop trying to describe your character with text. Show the AI what your character looks like, then tell it what to do.

Nano Banana Pro: Building Your Character Foundation

The key to consistency is creating a strong reference image first. This becomes your character's DNA. Every future image will reference this.

Start by using Nano Banana Pro to generate a side-by-side image—one image with a close-up face on the left and a full-body view on the right, both showing the same character. You need both the facial details and the full body design captured in a single reference. When you upload this later, the AI can see everything about your character at once.

Once you have that reference image, the workflow becomes simple but powerful:

1. Upload your character reference

2. Use the key phrase "featuring the same character shown in the reference image"

3. Describe the new scene, action, or context

4. Nano Banana Pro maintains facial features, clothing, and identity while adapting to the new environment

Nano Banana Pro supports up to 14 reference images (6 with high fidelity), allowing for "Identity Locking"—placing a specific person or character into new scenarios without facial distortion.

Veo 3.1: Start and End Frame Control

Where Nano Banana Pro handles character consistency across still images, Veo 3.1 brings that same control to video—with a game-changing feature: start and end frame specification.

Direct your story like a filmmaker: specify the exact starting and ending frames of your video to achieve complete narrative control.

Here's how it works:

The frames-to-video feature makes it possible to generate a video by providing the first and last frames. The model will then generate a video where the first and last frames are the provided frames.

You can now define both ends of the journey. Upload both the first and last frame, and the model will fill in the motion between them.

This means you can:

- Lock in your character's starting pose and ending pose

- Control narrative arcs with precision

- Create specific transformations or movements

- Maintain character identity throughout the video

Maintain perfect consistency across all your content. Upload reference images or videos to lock in a specific style, character, or aesthetic for unlimited generations.

The Power Combo: Use Nano Banana Pro to create your start and end frame images with perfect character consistency, then feed both into Veo 3.1 to generate the video between them. The workflow ensures the same character appears in both frames with identical facial features, hair, skin tone, and overall appearance, while only pose and expression change.

Why This Changes Everything

Before these tools, character consistency required one of two things:

1. Traditional animation workflows: Frame-by-frame control by human artists. Expensive. Time-consuming. Skill-intensive.

2. Expensive fine-tuning: Train custom AI models on specific characters. Requires technical expertise, GPU access, and thousands of training images.

Now? Upload a reference image, write a prompt, and generate the same character in different scenarios—whether it's a green alien riding a bike, taking selfies, or shooting hoops—while preserving key details like facial features, clothing, and overall style.

The applications are massive:

Comics & Sequential Storytelling: Generate the same character across multiple panels without the character morphing between frames.

Marketing & Branding: Keep mascots or avatars uniform across campaigns, ensuring brand consistency across dozens or hundreds of assets.

Video Content Creation: Content creators produce 1080p YouTube Shorts, Instagram Reels, and TikToks at scale while maintaining a signature style.

Game Development: Creating game assets with multi-image fusion lets you maintain consistent art styles across all your game characters.

The Technical Breakthrough

What makes this work isn't just better AI models—it's better control mechanisms.

The character consistency algorithm analyzes and extracts key identity markers from your reference images, including facial structure, distinctive features, color palette, body proportions, and style signature.

When generating new variations, the system preserves these core identity markers while adapting rendering rules to the target style (realistic, cartoon, anime, etc.). The result: a consistent character AI that stays recognizable across diverse artistic treatments.

For Veo 3.1, the start and end frame control adds temporal consistency to the mix. This feature is especially valuable for creators aiming to maintain visual consistency across scenes or produce longer, more cohesive videos. By anchoring the start and end points, Veo 3.1 can ensure smooth transitions.

The Professional Workflow

Here's the workflow that's emerging among professional creators:

Step 1: Generate your foundation character with Nano Banana Pro using a detailed prompt that captures all essential features.

Step 2: Create a side-by-side reference image (close-up + full body) that becomes your character's canonical look.

Step 3: For new scenes, upload the reference and use identity-locking language: "featuring the same character shown in the reference image. Keep all core design elements consistent."

Step 4: For video, use Nano Banana Pro to generate start and end frame images with your character in different poses.

Step 5: Feed both frames into Veo 3.1 with a prompt describing the motion, camera movement, and audio between them.

Step 6: Extend the video using Veo 3.1's extend feature, which takes an existing video and extends it based on the last frames consistently and seamlessly, making longer videos that remain consistent throughout.

The workflow uses 5 reference images to maintain character consistency. The workflow automatically handles character identity preservation across all generated content.

The Limitations (Yes, There Are Some)

This isn't magic. There are constraints:

Veo 3 does not allow using a first frame and reference images simultaneously. Users must choose either a first frame (optionally with a last frame) or a set of 1-3 reference images, but not both in the same generation request.

Videos start and end at the given frames, but they don't always transition naturally between the two. There's still a lot of room for improvement in the model.

And character drift can still happen if you're not careful. To stop face drift across a series of images, use a stable identity tag near the start of the prompt, keep the same viewpoint (like 3/4 mid-shot), and fix a minimal style stack. Compare every new output to your saved anchors and repair small shifts with targeted inpainting.

What's Next?

The real opportunity here isn't just technical—it's creative. For the first time, individual creators can produce character-driven content at scale without animation studios, without expensive equipment, and without years of technical training.

Bring your stories to life with character consistency technology. Whether you're crafting webcomics, children's books, or game narratives, our AI image generator remembers your characters' unique traits across every frame. Create entire visual stories where heroes, villains, and supporting cast stay perfectly recognizable from panel to panel.

We're watching the early days of AI-native storytelling emerge. The creators who master these tools now—who learn to build reference libraries, control start and end frames, and chain sequences together—are building unfair advantages.

Because the barrier to character-driven content just collapsed. And most people haven't noticed yet.

BTW: Input the initial frame photo, Nano Banana can generate subsequent frame photos, which you can edit as needed to create a movie storyboard that meets your needs. Next, use the picture-to-video function to generate a coherent video from each frame without changing the characters, for the production of a completely high-quality movie. We're not far from solo creators producing full animated series from their laptops.

Have an ad concept ready but don't want to deal with expensive shoots or stock footage? ScriptKit lets you generate, curate, and edit AI ads in one platform.

What ScriptKit gives you

Give ScriptKit a shot — go from concept to finished AI ad without wrangling teams or gear.

The Recap: Brazil's consumer protection agency ordered Meta to immediately suspend its policy banning third-party AI chatbots from WhatsApp, threatening daily fines if Meta doesn't comply within five days. The regulatory intervention forces Meta to open WhatsApp—its most valuable communication channel with over 2 billion users—to competing AI assistants, directly threatening the closed-ecosystem AI moat Meta has been trying to build. Brazil's move follows similar regulatory pressure in other markets where governments are using antitrust and consumer protection laws to force platform interoperability, revealing that Big Tech's AI strategy of locking users into proprietary assistants may not survive regulatory scrutiny.

Unpacked:

Bottom line: Brazil's order to open WhatsApp to rival AI chatbots exposes the fragility of Big Tech's platform-based AI strategies in an era of aggressive regulation. Meta's plan to leverage its 2 billion WhatsApp users as exclusive distribution for Meta AI collapses if every major market can force interoperability through consumer protection law. The real story isn't one regulatory decision in one country—it's the global template this creates for dismantling closed AI ecosystems. As companies pour billions into AI development and infrastructure, regulators are rewriting the rules on who gets to distribute AI and how. Meta's challenge isn't just building competitive AI models—it's defending the right to use its own platforms as exclusive distribution channels.

Meta launches its own AI infrastructure initiative betting that controlling energy and data center capacity will prove more defensible than model competition alone, as the AI race shifts from algorithms to physical infrastructure.

Apple and Google's Gemini deal reveals that even the most vertically integrated tech giant sees AI as a strategic outsource rather than core capability, collapsing the narrative that proprietary foundation models create sustainable differentiation.

Slackbot becomes an AI agent as Salesforce turns its most-used communication platform into an orchestrator, attempting to capture the workflow layer before competitors can lock in enterprise AI behaviors.

OpenAI buys tiny health records startup Torch for reportedly $100M, revealing the real bottleneck isn't model power but enterprise data access as AI labs spend heavily on domain-specific integration capabilities.

Microsoft scrambles to quell community fury around new AI data centers, exposing a new vulnerability for Big Tech where environmental backlash could constrain the energy capacity needed to scale AI competitively.

Deepgram raises $130M at $1.3B valuation and buys a YC AI startup, consolidating the speech-to-text infrastructure layer and suggesting specialized AI infrastructure creates more durable moats than general-purpose models.

Anthropic announces Claude for Healthcare following OpenAI's ChatGPT Health reveal, as both leading labs race to verticalize simultaneously—indicating industry-specific applications, not general models, are the path to defensible revenue.

Amazon buys Bee AI wearable betting that wearables represent a new interface layer for capturing consumer data and attention that could shift power away from phones as the next AI frontier focuses on control of human-AI interaction points.

Tech unicorns surpass 100 new companies in 2025 as the proliferation of AI unicorns suggests the market is betting on specialized vertical solutions over generalist platforms, contradicting the winner-take-all narrative around foundation models.

Best of AI™ Team

Was this email forwarded to you? Sign up here.