{"id":21270,"date":"2026-07-07T00:45:41","date_gmt":"2026-07-06T23:45:41","guid":{"rendered":"https:\/\/innovatenews.site\/index.php\/2026\/07\/07\/i-replaced-adobe-figma-and-canva-with-chatgpt-and-one-open-source-tool\/"},"modified":"2026-07-07T00:45:49","modified_gmt":"2026-07-06T23:45:49","slug":"i-replaced-adobe-figma-and-canva-with-chatgpt-and-one-open-source-tool","status":"publish","type":"post","link":"https:\/\/innovatenews.site\/index.php\/2026\/07\/07\/i-replaced-adobe-figma-and-canva-with-chatgpt-and-one-open-source-tool\/","title":{"rendered":"I replaced Adobe, Figma, and Canva with ChatGPT and one open-source tool"},"content":{"rendered":"
\n

The topic I replaced Adobe, Figma, and Canva with ChatGPT and one open-source tool<\/strong> is currently the subject of lively discussion \u2014 readers and analysts are keeping a close eye on developments.<\/p>\n

This is taking place in a dynamic environment: companies\u2019 decisions and competitors\u2019 reactions can quickly change the picture.<\/p>\n

I contemplate turning my life around every time I watch a movie or TV show with a career-oriented protagonist. Gilmore Girls inspired me to want to be a journalist (manifesting works), Brooklyn 99 had me convinced I’d make a great detective, and How to Get Away with Murder convinced me a lawyer was what I was meant to be. The only career no amount of good TV or movies can talk me into is anything to do with art. I’m a terrible artist.<\/p>\n

My handwriting, even on its good days, looks like it was written during an earthquake, and the most I can draw are stick figures that somehow still come out anatomically incorrect. It’s unfortunately the same story when it comes to design tools. Despite the fact that I enjoy designing slide decks and ideating for them, I’m not very good at it. That’s why I’ve always leaned on templates to do the heavy lifting, and I occasionally tinker around with the fancier editing apps too. However, this open-source design tool paired with ChatGPT has done more for my design output than Adobe, Figma, and Canva combined.<\/p>\n

\"\"<\/figure>\n

If you’ve been following the major AI labs, you know well that there’s a pattern to how products are launched. When one company launches a chatbot, you’ll find every other lab shipping their own within the month. When AI browsers became the thing, suddenly everyone had one in the works. When terminal-based coding agents like Claude Code took off, every lab scrambled to put out their own. The same wave is now hitting design tools. A few weeks ago, Anthropic announced their newest product under research preview called Claude Design, and the idea behind it is that you describe what you want and get back a finished design instead of instructions on how to design it yourself.<\/p>\n

While a lot of AI tools have offered the “describe, and we’ll generate” functionality, like NotebookLM’s Slide Decks and even image models like ChatGPT’s and Gemini’s, what sets Claude Design apart is that you aren’t stuck with whatever it generates. You can edit elements directly, refine through chat, and export to PDF, PPTX, or Canva. However, Claude Design is expensive, and you’re limited to Anthropic’s own models. You can also only access it if you have a Claude Pro subscription or above. Fortunately, the open-source community found a solution, as always, and built an open-source, self-hostable clone called Open Design that does the same thing without tying you to Anthropic’s models or a paid plan.<\/p>\n

Open Design is the nexu-io team’s take on all this, released under the Apache-2.0 license and built to run entirely on your own setup. It keeps Claude Design’s describe-build-iterate loop but swaps out the closed model for whatever agent you’ve already got, since it runs on a BYOK model.<\/p>\n

It auto-detects whatever coding agent you’ve already got installed and signed into, whether that’s Claude Code, Codex, Gemini CLI, or OpenCode, and wires it straight into the design workflow. This means you don’t need to get an “Open Design” subscription for it or plug in an API key for a cloud or local LLM (though you can), and you can simply lean on a coding agent you’ve already logged in to. In my case, that’s Codex.<\/p>\n

\"\"<\/figure>\n

Getting started with Open Design is incredibly simple, too. While the process of getting it running on your own local server is a bit trickier, it has a desktop app for anyone who’d rather not touch a terminal. Beyond that, you can install it as an MCP server inside your agent, run it in Docker, or clone the repo and run from source like I did initially. I now use the desktop app exclusively, and I’ve had no complaints.<\/p>\n

I’ve been using Open Design for a few days now, and not once have I opened Canva, Adobe, or Figma for anything I’d typically open them for. Open Design has been the app I go straight to, and it’s because of the features it offers and the quality of the outputs. Since it’s using OpenAI’s models via Codex, the actual design work rides on whatever model Codex is pointed at, and the results have held up well across everything I’ve thrown at it. For instance, I needed to make a slide deck recently for a presentation for which I already had the script ready. Instead of heading to Canva or Adobe Express, picking a template, and spending two hours tweaking elements myself, I opened Open Design and gave it a prompt. I pasted my entire script and asked it to generate a slide deck for me.<\/p>\n

Instead of just generating the presentation right away, it asked for my preferences first. It asked for the intended audience, the number of slides I wanted, if I had a specific design system or brand in mind, and if I had any instructions. Only after I answered did it start building. The deck was ready within minutes, and the output was good to go without any editing needed! However, on the rare occasion I did want a change, editing was as simple as selecting the element and telling it what to fix (or fixing it myself).<\/p>\n

Once I’m done working on my deck, I can download it as a PDF, editable PPTX, image, zip file, or as a standalone HTML. As someone who is tired of generic slide decks that scream templates and is good at directing models to do what I want, Open Design fits the workflow I’ve wanted for years without asking for a separate subscription.<\/p>\n

Excuse the clich\u00e9, but all that glitters is certainly not gold. The deck from my example above burned through 12% of my Codex session limit and 1% of my weekly limit in one go, which means I haven’t really escaped rationing \u2014 I’ve just swapped Claude Design’s weekly cap for Codex’s. However, there’s a way around it: Open Design lets me switch to another model or a local LLM, so I can simply switch if needed.<\/p>\n