This week in AI, the tools that power creative work got a lot more accessible. From Canva letting any of its 265 million users spin up a polished interactive website from a plain-language prompt to Adobe pushing autonomous agents inside Photoshop and Premiere, the gap between idea and finished product keeps narrowing for independent makers and small creative businesses. Four stories worth paying attention to.
Canva Code 2.0: AI Website Building for Every User, Including Free Accounts
Canva launched Code 2.0 this week, opening its AI-powered site and app builder to all 265 million monthly users regardless of plan tier, free accounts included. Describe what you want in plain language and Canva builds a fully interactive, responsive experience; you can then drag-and-drop elements, import HTML from other tools, and publish without writing a single line of code. Generation speed is 75% faster than the original Code tool and 50 new templates cover common small-business use cases from event pages to product showcases. For makers running an Etsy shop or a commission-based side hustle, this removes the last real excuse for not having a polished standalone web presence.
Adobe Puts AI Agents Inside Photoshop, Premiere, and Illustrator
Adobe expanded its creative agent technology into public beta across Premiere, Photoshop, Illustrator, InDesign, and Frame.io. A sidebar in each app accepts plain-language instructions and the agent executes multi-step production tasks: sorting and renaming media in Premiere, batch background removals in Photoshop, and rolling brand updates across entire InDesign layouts in one pass. For one-person studios and small creative teams, this is the kind of automation that previously required either a bigger team or hours of manual work. The same agent extends into ChatGPT, Microsoft Copilot, and Claude, so it works alongside whatever platform you already run your projects in.
Krea 2: A Free Professional-Grade Image Model You Can Run and Fine-Tune Yourself
Krea released the weights of Krea 2, its 12.9-billion-parameter diffusion model, as free open weights on Hugging Face with a license that allows commercial use for teams under 50 seats. Two versions ship: Krea 2 Raw, built for fine-tuning and LoRA training so you can train it on your own style or subject matter, and Krea 2 Turbo, an 8-step distilled model that generates 2K images in roughly two seconds on consumer hardware. A key distinction from other open releases is that Krea trained on real images only, removing synthetic data from the mix to avoid quality ceilings. For independent creators and small studios, this means a professional-grade image generator you can run, customize, and own rather than pay per query to a closed API.
ByteDance Seedream 5.0 Pro Exports Your Graphic as 10 Ready-to-Edit Layers
ByteDance launched Seedream 5.0 Pro, an image model built for professional graphic production workflows rather than simple generation. Prompt it with a complex layout and it outputs the finished image plus 10 or more independent transparent-PNG layers that open natively in Photoshop or Figma, with background areas auto-filled under previously hidden subjects so you skip masking entirely. The model also renders legible text in 14 languages inside the image, including right-to-left scripts like Arabic, which opens up localized poster and merch designs without a second editing pass. It runs on third-party platforms including fal and ComfyUI so access is not limited to ByteDance's own apps.
That's Tech Thursday
That's the tech and AI news that actually matters for creators this week. Want it turned into something physical? That's what the studio does. See you next Thursday.
Some links in this post are affiliate links. Prices stay the same for you, I earn a small commission if you buy through them.