In partnership with
Shoppers are adding to cart for the holidays
Peak streaming time continues after Black Friday on Roku, with the weekend after Thanksgiving and the weeks leading up to Christmas seeing record hours of viewing. Roku Ads Manager makes it simple to launch last-minute campaigns targeting viewers who are ready to shop during the holidays. Use first-party audience insights, segment by demographics, and advertise next to the premium ad-supported content your customers are streaming this holiday season.
Read the guide to get your CTV campaign live in time for the holiday rush.
OpenAI Just Turned Image Generation Into Infrastructure
Listen—GPT Image 1.5 dropped yesterday, and everyone's talking about "4x faster!" and "better text rendering!" But here's what actually matters: OpenAI just repositioned AI image generation from a creative toy into production infrastructure.
The surface story is simple. OpenAI accelerated their launch from January to December 16th because Google's Nano Banana Pro was eating their lunch. Google had multilingual text, 4K output, provenance tracking. Classic competitive pressure, right?
But OpenAI didn't copy those features. They optimized for what enterprises actually need: instruction adherence, iterative control, and economic efficiency. And the results are dominating every benchmark—LMArena, Design Arena, Artificial Analysis. They're ranked #1 across the board.
Here's the kicker: Benchmark wins don't tell the whole story. GPT Image 1.5 outputs look like commercial photography—polished, professionally lit, visibly controlled. Nano Banana Pro produces images that feel more like candid photographs, the kind of authentic aesthetic creators often prefer. The benchmarks measure "how well does it follow my brief," but they don't measure "does this look like something a human actually made."
The real breakthrough? Text rendering that actually works. For two years, AI generators gave you "Bleee Noote" instead of "The Blue Note." Designers were stress-testing GPT Image 1.5 with magazine covers, vintage jazz posters, infographics with data tables—and the typography just works. That single shift moves this from "art generator" to "design assistant."
Then there's mask-based editing. You can now change a person's outfit without the model rewriting their face, the lighting, and the background. E-commerce teams can generate one product shot, then iterate through twenty different backgrounds without regenerating the product itself. Marketing teams can batch-generate localized campaigns—different text overlays, character variations—all consistent with brand guidelines.
The economics matter too: 20% cheaper than GPT Image 1, but with hidden costs in reasoning tokens the model uses internally to plan layouts. At scale though, the 4x speed compounds—you iterate more in the same session, reducing total cost-per-final-output.
What this reveals: OpenAI is moving upmarket. This isn't Midjourney's "fun for artists" play. This is "enterprise visual asset infrastructure." If you're generating 1000+ images monthly, the production efficiency advantage is real. If you're a solo artist seeking creative surprise, Nano Banana Pro remains the better choice.
The visual generation arms race just accelerated.