🚨Sequoia is joining Anthropic's $25 billion funding round despite already backing OpenAI and xAI, shattering the decades-old VC taboo against investing in direct competitors as the biggest firms abandon winner-take-all strategies in AI.

Connect your email, and you’ll instantly get a CRM with enriched customer insights and a platform that grows with your business.

With AI at the core, Attio lets you:

Join industry leaders like Granola, Taskrabbit, Flatfile and more.

👉 Try Attio Pro for free

AI-generated video has a reputation problem. The hands morph, the physics break, and viewers can spot the uncanny valley from a mile away. But a new workflow is emerging that's making AI video ads look... well, real.

The secret? It's not about throwing a text prompt at a video model and hoping for the best.

How it works:

Instead of generating video in one shot, this workflow treats video like what it actually is: a sequence of individual frames that need to maintain consistency and realism.

Step 1: Start with a photorealistic image

The foundation is using JSON prompts for the initial image generation. JSON-structured prompts give AI image models precise, parameter-level control over composition, lighting, camera angles, and subject positioning—resulting in images that look like they came from a DSLR, not a neural network.

Think: camera metadata, lighting conditions, exact positioning coordinates, lens specifications.

Step 2: Use that image as the start frame

Once you have a photorealistic base image, it becomes the first frame of your video. This anchors the entire sequence in reality.

Step 3: Generate each subsequent frame using the previous frame

Here's where the magic happens: instead of generating a 5-second clip all at once (which compounds errors and creates drift), you generate frame-by-frame or in very short bursts, using each completed frame to inform the next.

This approach maintains visual consistency across the sequence and prevents the AI from wandering into surreal territory.

Step 4: Sync to script cadence

The breakthrough insight? Script pacing matters as much as visual quality. The number of words and syllables per section determines how long each visual segment needs to hold, which affects how many frames you need to generate and how much motion can occur between them.

A script section with 15 syllables at conversational pace = ~1.5 seconds of video = ~36 frames at 24fps. Map your script's rhythm to your frame generation workflow, and suddenly your visuals and voiceover lock together seamlessly.

Why this matters now:

Traditional video production for ads is expensive. A professional 30-second commercial can cost $10,000-$200,000+ when you factor in actors, crews, locations, and post-production.

AI video tools promised to democratize this... but the early results were unusable for brands that care about quality. The tech could make weird, dream-like clips, but not the crisp, believable product demos or testimonial-style ads that actually convert.

This frame-by-frame, JSON-prompted workflow bridges that gap. It's still being dialed in—the workflow isn't one-click yet—but early adopters are getting results that pass the "scroll test" on social feeds..

What's next:

The workflow is still in the optimization phase. Creators are experimenting with:

- Automated JSON prompt generation from simple text descriptions

- Script-to-frame calculators that automatically determine optimal segment lengths

- Batch processing tools that handle frame-by-frame generation without manual intervention

- Style consistency models that ensure frame N+1 doesn't drift from frame N

The goal? Make creating a photorealistic AI video ad as easy as writing a script and clicking generate.

We're not there yet—but the fact that it's possible at all would have seemed like science fiction 18 months ago.

BTW: The human eye can detect visual inconsistencies in as little as 3-4 frames (1/8th of a second). That's why frame-by-frame generation with anchored start frames works better than single-shot generation—it prevents micro-inconsistencies from compounding into the uncanny valley effect that makes viewers instinctively distrust AI-generated content.

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: Netflix has revised its acquisition offer for Warner Bros to all-cash terms as it races to block a potential Paramount deal, signaling that streaming platforms are willing to deploy massive capital to secure premium content libraries even as AI-generated content emerges as a potential threat. The move demonstrates that owning marquee franchises and established IP remains a defensible moat in an era where technology alone can't replicate decades of audience attachment and cultural cachet.

Unpacked:

Bottom line: Netflix's all-cash Warner Bros bid signals that even in the AI era, owning irreplaceable content libraries trumps technology advantages. The company is betting that decades of audience attachment to Batman, Harry Potter, and HBO prestige shows creates moats that generative AI can't breach—a direct rejection of the thesis that AI will commoditize all content. If the deal closes, it reshapes streaming competitive dynamics by creating a platform with both Netflix's recommendation algorithms and Warner Bros' cultural franchises, forcing competitors to respond with their own acquisitions or accept permanent content disadvantage.

VCs are betting big on AI security as rogue agents and shadow AI create compliance nightmares, establishing a new multi-billion dollar category around controlling employee AI tool adoption that organizations can't ignore.

OpenAI's 2026 focus is "practical adoption" according to COO Sarah Friar, acknowledging the fundamental gap between AI capability and business utility that's forcing a strategic pivot from raw model performance to implementation economics.

BioticsAI gained FDA approval for its AI-powered fetal ultrasound product, proving regulatory approval is now the real bottleneck for AI adoption in healthcare rather than technology maturity.

Threads surpassed X with 141.5 million daily mobile users, demonstrating Meta's ability to cross-promote and iterate faster than competitors is consolidating social media into a single dominant platform.

Amazon CEO says tariffs are starting to "creep into" pricing as supply chain arbitrage buffers deplete, forcing e-commerce leaders to choose between margins and market share in a structural shift affecting tech valuations.

Sony's TV business is being taken over by TCL, showing how hardware commoditization forces legacy consumer electronics players into secondary roles as AI and software shifts permanently cede entire categories.

Developers question whether agentic coding actually works, revealing a massive gap between VC-hyped claims and real-world productivity gains that threatens the narrative justifying AI infrastructure spending.

Porsche sold more electrified cars than pure gas-powered cars in Europe in 2025, demonstrating luxury automakers are hitting EV adoption tipping points faster than mass-market competitors.

Best of AI™ Team

Was this email forwarded to you? Sign up here.