Introducing Runable 2.0: The Best Way to Work with AI
Three months ago, we launched Runable. Over 500,000 people signed up. They built websites, created slide decks, ran deep research, generated videos all from a single prompt.
But as we watched how people actually used it, one thing became clear: the biggest problem wasn't what Runable could do. It was that Runable started doing things before it fully understood what you needed.
You've felt it.
You type a prompt. The AI goes off and builds something. You wait. What comes back is close but not quite right. So you tweak and try again. And again. And again.
That's not a prompt problem. That's an agent problem .
That's not how great work gets done.
Runable 2.0 fixes this at the root. And while we were at it, we shipped Canvas Mode, smarter slides, full-stack websites, and a built-in AI voice agent.
Here's everything that changed and why it matters to you.
The Numbers First
Runable 2.0 scores 92.1%, 68.3% , 86% and 76% on GAIA , DRACO, SlidesBench and BrowserComp respectively, one of the hardest real-world AI performance tests. That puts us among the best AI agents in the world. Better research, better outputs, fewer wrong turns.
What Changed: The New Agent asks, plans and then builds it.
Before (v1) :
You typed a prompt → Runable built → You hoped it matched what you meant.
Now (v2):
Here's what actually happens now:
It asks you first - Before touching anything, Runable shows you a short set of questions your audience, your style, what you need included.
It shows you the plan - After your answers, Runable shows you a visual preview of what it's about to build the structure, layout, content outline before generating anything. You approve it or adjust it. Only then does it build.
You edit in place - Once it's built, ask AI to change anything right inside the output.
Want to change a headline ? Swap a color ? Rewrite a section ? Ask the AI directly inside the output. No copy-pasting. No switching tools. No new chat.
Rollback anytime. Didn't like a change ? Go back to any previous version in one click.
Fork and try multiple directions - Want to test 3 different website themes ? Fork the chat, try all three in parallel, keep the best one.
Slides: Visual-First, Editable, Actually Yours
Before (v1):
Runable generated slides from your prompt. Content was decent, but design felt generic. To change a layout or color scheme, you had to regenerate or go to another tool.
Now (v2):
Runable asks about your audience and tone first, shows you the slide-by-slide structure before generating, then builds a deck that matches what you actually described.
New: Nano Banana Mode - a visually bold slide style for creative presentations.
Every slide is editable inline . Tell the AI what to change - it changes it. You never leave the window.
Websites: Not a Demo. A Real, Deployed Product.
Before (v1):
Runable built websites fast but for anything serious (payments, database, a real domain), you had to leave and set it up yourself.
Now (v2):
Every website Runable builds is a fully deployed, working product:
- Live URL - publicly accessible the moment it's built
- Database - store users, form submissions, anything
- Stripe - accept payments out of the box
- Custom domain - connect yours directly
- SEO + Analytics - built in, zero config
- Analytics dashboard - see who's visiting and what they're doing
- Version history - roll back your site just like a chat message
The agent asks what your site needs to do before writing a single line.
You answer once. It builds correctly the first time.
Canvas Mode: One Place for Every Visual
Before (v1):
Images and video lived in separate flows. Generate in chat, download, open Canva or Photoshop to edit, come back, repeat.
Now (v2): Canvas Mode
is a single visual workspace generate, edit, and enhance images and videos without leaving Runable.
- Generate images - Nano Banana Pro, Seedream 4.5, FLUX.1 Kontext, and more
- Generate videos - Google Veo 3.1, Kling 3.0 Pro, Sora 2 Pro, Seedance 1.5 Pro
- Mark Edit - paint over any part of an image, describe the change, done
- Motion Control - animate between two images with smooth transitions
- AI Effects - cinematic, stylized, creative — for both images and videos
- Virtual try-ons - show a product on a real person.
No Photoshop. No Premiere. No Canva.
One canvas.
AI Voice: Your Website Can Now Listen
Before (v1):
Websites built with Runable were static, forms, text, buttons. No way for a visitor to talk to your site.
Now (v2):
You can add an AI voice agent to any website you build. Visitors speak the AI listens, understands, and responds in real time.
Use it for:
- Feedback - visitors speak instead of typing
- Support - answers questions 24/7, based on your content
- Lead qualification - the agent asks and captures responses
- Product walkthroughs - guide users through your product conversationally
The agent knows your site's content and responds in the tone you define.
The Full Picture: What You Now Get with Runable 2.0
| What's New | What It Means for You |
|---|---|
| Agent asks first | No more wrong outputs, it understands before building |
| Visual plan preview | See what you're getting before it's generated |
| Inline editing | Change anything by asking AI, right inside the output |
| Rollback | Go back to any version in one click |
| Chat forking | Try multiple directions in parallel, risk-free |
| Canvas Mode | Generate + edit images and videos in one visual workspace |
| Nano Banana Mode | Bold, creative slide style, not a corporate template |
| Full-stack websites | Live URL, database, Stripe, domain, SEO, analytics, all included |
| AI Voice Agent | Add a talking, listening AI to any website you build |
Try Runable 2.0
Every feature in 2.0 came from watching 500,000 people use v1, what frustrated them, where they had to switch tools, what they wished AI could just handle.
2.0 handles it.
Questions ? support@runable.com
