🔥Ugly deck? Fix it with AI

Blog

← Blog

How CubeOne Generates Full AI Slide Designs in Seconds

Apr 21, 2026


Most AI slide tools ask you to craft a careful prompt and then wait a minute or more for a slide to render. CubeOne takes a different approach: attach a style reference, drop a sentence, and the slide renders in seconds. Every edit after that regenerates in about 2 seconds.

This post walks through exactly how that works, and why CubeOne can be this fast without cutting quality.

Video walkthrough

In this post:
  • → Start with a style reference (or skip it)
  • → Auto-extracted style guide: colors and style prompt
  • → Generate a designed slide in seconds
  • → Add another slide in one sentence
  • → Regenerate in about 2 seconds
  • → Why CubeOne is faster than other AI slide tools

1. Start fast: attach a style, drop a sentence

On the home screen you get a single input. Attach any image you want the deck to look like (a screenshot from Dribbble, a brand asset, an existing slide you love) and write what the slide is about. That is the full setup.

Attach a style reference image and drop a sentence
CubeOne home input with a style reference image attached and a one-line prompt

If you do not care about a specific brand or aesthetic, just skip the reference. CubeOne will pick a style that fits the content on its own.

2. Style guide, extracted automatically

After a few seconds of processing, you get a style guide. CubeOne pulls the dominant colors from your reference image and writes a style prompt that captures the aesthetic in words: typography, grid, mood, contrast, accent behavior. You can edit any of it, or leave it alone and move on.

Colors extracted from your image plus an auto-written style prompt
Brand panel showing extracted color palette, an auto-generated style prompt, and the original reference image

3. Hit generate, slide appears in seconds

Click generate. A designed slide renders in seconds. You can see the exact style prompt CubeOne used on the right panel, along with the original reference image attached at the bottom. The output genuinely looks like the reference: same color system, same weight of type, same visual energy.

Finished slide with prompt and style reference visible
Finished slide reading THE FASTEST AI SLIDE MAKER PERIOD on a blue background, with the prompt and style reference shown in the right panel and Gemini 3.1 Flash Lite as the active model

If something is off, just add what you want changed to the prompt on the right and hit regenerate. The next version appears in about 2 seconds.

4. Add another slide in one sentence

Adding a new slide is the same story, not a different story. No drag-and-drop boxes like PowerPoint. No careful prompt engineering to coax the right result out of a slow model. Click Add Slide, type what you want, optionally attach a file, hit generate.

New slide: “Add a thank you page with attached QR code, blue background.”
Empty new slide with a short prompt on the right and a QR code attached as a file

In about 2 seconds:

Generated thank-you slide with the attached QR code
Rendered thank-you slide with a large QR code on a blue background

5. Iterate at 2-second speed

Want the QR code bigger? A link in the footer? Less visual clutter? Add the note to the same prompt box and hit regenerate. The slide reflects your feedback almost immediately. No restart, no reupload, no reprompting from scratch.

Add follow-up notes at the end of the prompt and regenerate
Slide 2 regenerated as SCAN TO TRY with the added feedback highlighted at the bottom of the prompt box

Manual edits work alongside AI regeneration. You can click directly into text, swap images, or reorder slides, which is often the fastest option for small cosmetic fixes.

How CubeOne can be this fast

The short answer: CubeOne owns the design system, not just the prompt.

Most AI slide tools are wrappers. They take your input, pass it to a large general-purpose model, and ask that model to figure out layout, typography, hierarchy, color, and composition all at once. That is why they need expensive models and minute-long generations, and why the output often still looks generic.

CubeOne is built around design itself. Context retrieval, style awareness, asset placement, layout logic, and typographic hierarchy all live inside CubeOne's system. The model is only responsible for the last mile: writing the final slide code under a tight, well-specified brief.

That lets CubeOne run on Gemini 3.1 Flash Lite, the lightest, cheapest, fastest model in the lineup, and still produce designs that beat other tools running on the most expensive models available. Less latency per slide, lower cost per slide, better output per slide.

FAQ

How long does CubeOne take to generate a slide?
The first slide renders in a few seconds after you hit Generate. Regenerating a slide after an edit takes around 2 seconds. Generating a full deck scales with slide count, but each slide stays in the seconds range rather than the minute range many AI slide tools require.
How does CubeOne stay efficient while producing polished designs?
The design intelligence lives in CubeOne's system (layout, typography, style transfer, context retrieval, asset awareness), not inside a large general-purpose model. The model only writes the final slide code under a tight brief. That means less compute per slide and a better visual result per slide.
Why is CubeOne faster than other AI presentation tools?
CubeOne owns the design layer. Other AI slide tools wrap a generic large model around a prompt template and wait for it to do everything, including layout, typography, hierarchy, and color. CubeOne's system handles layout, style transfer, context retrieval, and asset awareness directly. The model only needs to write the final slide code, so it can be a light, fast model (Gemini 3.1 Flash Lite) and still produce designs that outperform competitors running the most expensive models.
How do I tell CubeOne how many slides to generate?
You say it in the prompt. For example: "slide number: 10" or "make a 6-slide pitch deck." After the initial deck is generated, you can still manually add, move, duplicate, or remove slides from the sidebar at any time.
Set slide count directly in the prompt
Home input with slide number 1 highlighted as part of the prompt
Do I need to write a long, carefully engineered prompt?
No. One sentence is usually enough. CubeOne fills in the missing structure, style, and layout automatically. If you want tighter control, attach an image as a style reference and CubeOne will match your aesthetic without you having to describe it in words.
How fast can CubeOne regenerate a slide after an edit?
About 2 seconds. You add your change to the prompt box on the right (for example, "make the QR code bigger" or "add a link in the footer"), hit Regenerate, and the slide updates almost immediately.
Which AI model does CubeOne use?
CubeOne uses Gemini 3.1 Flash Lite by default. It is fast, affordable, and reliable. CubeOne's internal design system does the heavy lifting around style, layout, and context, so a lightweight model is more than enough to produce premium results.
Can I manually edit a slide after the AI generates it?
Yes. You can edit text inline, replace images, or reorder slides manually at any time. AI generation and manual editing coexist, so you never get stuck waiting on AI for a small change.

Try It

If you have spent a full minute waiting on another AI slide tool to render a slide you are probably going to throw away, CubeOne is the fix. Attach a reference, type one sentence, watch the slide appear.