Kimi Slides Guide (2026): Free AI Presentations With Charts You Can Actually Edit

Every AI presentation tool promises a deck in sixty seconds. Most deliver slides where the "chart" is a picture of a chart — so when your CFO asks to change one number, you rebuild the whole thing by hand.

Kimi Slides is the one that fixed this. It renders charts and diagrams as native objects, not images, and exports real .pptx — so the bar chart lands in PowerPoint as an actual PowerPoint chart, with the data still attached. Change a figure, it redraws.

This Kimi Slides guide covers what it is, exactly what's free versus paid (including a pricing situation nobody's writing about), the full walkthrough, my copy-paste prompts, and an honest comparison with Gamma and Canva. If you came from my reel — this is the whole thing.

What Kimi Slides Actually Is

Kimi Slides is the presentation agent inside Kimi, the assistant built by Moonshot AI. Not a separate app — a mode you select inside the chat, alongside Deep Research, Docs and Sheets. That's Moonshot's direction: one agent that does knowledge work end to end rather than a folder of disconnected tools.

The official docs describe it as "an intelligent presentation generation tool launched by Kimi, powered by the K3 large model." Worth flagging, because most guides you'll find still say K2 — this thing moves fast enough that month-old articles are wrong.

The workflow is agentic rather than templated. Give it a topic, a pile of documents, or an image, and it researches, writes an outline, generates content and designs the layout as a sequence of steps you can watch. It takes PDF, Word, Excel, PPTX, Markdown, TXT and images as input, and exports to PowerPoint, Google Slides or PNG.

For visuals it uses Google's Nano Banana Pro image model to generate custom infographics and illustrations instead of pulling stock art — which is why decks from its Visual mode don't look like everyone else's.

The Two Modes: Adaptive vs Visual

This is the fork in the road, and picking wrong wastes an hour.

	Adaptive	Visual
Built for	Research-heavy, content-first decks	Design-led, visually striking decks
Time to deck	~30–60 minutes	~5–10 minutes
What it does first	Reads and researches, then designs	Designs, using Nano Banana Pro visuals
Strength	Depth, citations, structure from your sources	Custom illustrations, infographic quality
Available on	Free tier (Adagio)	Paid tier (Moderato)

Adaptive is the workhorse. It behaves like an analyst: gathers material, structures an argument, then dresses it. Slow, but the content is the point.

Visual is the showpiece. It leads with design and generates original graphics rather than laying out text. Fast, and it looks it — in a good way.

Practical rule: Adaptive when the deck has to be right, Visual when the deck has to land. A market analysis for an investor gets Adaptive. A creative pitch for a client gets Visual.

What's Actually Free (No Hedging)

Here's the section you came for, straight:

The free plan is called Adagio. It costs $0 and it includes Kimi Slides in Adaptive mode, plus the K3 model and Deep Research. Not a trial. Not watermarked. Real decks, real .pptx export.

Two honest limits:

1. It's credit-metered, not unlimited. Free accounts get a small monthly pool of agent credits shared across everything Kimi does. A deck is a meaningful chunk of it. Think a handful of presentations a month, not unlimited. Check your credit balance before starting something big.
2. Visual mode is paid. Nano Banana Pro visuals require Moderato, listed at $19/month (roughly 20% less annually), which also brings 60 agent credits, priority speed and concurrent tasks. Above it sit Allegretto ($39), Allegro ($99) and Vivace ($199).

And now the part almost nobody has updated: on 19 July 2026, Moonshot paused new paid subscriptions because K3 demand pushed them to their compute limits. As of mid-August 2026 the paid tiers still can't be purchased. Existing subscribers keep everything; Moonshot says spots reopen in batches, with no date announced.

So the real answer to "should I pay for Visual mode?" today is: you probably can't. Which makes the free tier's Adaptive mode not the budget choice but the actual product for most new users. Plan accordingly — and since all of this moves, treat these prices as verified on 12 August 2026 and check kimi.com before budgeting.

Step-by-Step: Your First Deck

Step 1 — Sign up and open Slides

Go to kimi.com, create a free account, and look at the row of agent modes under the chat box: Swarm, Slides, Deep Research, Websites, Docs, Sheets, Design. Click Slides. No credit card, no separate signup.

Step 2 — Choose your mode

Pick Adaptive or Visual per the table above. On a free account Adaptive is your lane. Set expectations now: this runs for 30–60 minutes and you can walk away — it's a background job, not a chat reply.

Step 3 — Prompt it, or feed it documents

Two paths, and the second is better. Either write a detailed brief (my prompts are below), or upload your source material — the quarterly PDF, the Excel of transaction data, the research doc. Uploading is the single biggest quality upgrade available, because the deck gets built from your real numbers instead of the model's recollection of the internet.

You can also upload a PowerPoint template to steer the styling, though read the limits section before you count on it for strict brand work.

Step 4 — Watch it work

You get a visible plan: research, outline, content, design. Useful, because you can catch a wrong-shaped outline early rather than after fifty slides. Free accounts may queue at peak hours — evenings in Asia are the crunch.

Step 5 — Review and edit in the browser

Read it critically, especially every number. Edit text inline, change styling, regenerate individual slides you don't like. Don't regenerate the whole deck over one bad slide.

Step 6 — Export and check the charts

Export as PowerPoint (or Google Slides / PNG). Then do the test that matters: open the .pptx and double-click a chart. If PowerPoint opens its data grid, you have a real chart object and the deck is yours to maintain. That's the moment the tool proves itself.

My Prompts That Produce Good Decks

Vague prompt, generic deck. Every one of these specifies audience, structure and evidence. Copy and swap the details:

1 — The quarterly business review

Build a 12-slide Q3 business review for a Dubai real estate brokerage. Sections: performance vs targets, lead sources and cost per qualified lead, pipeline by community, what worked, what didn't, Q4 plan with three priorities. Use bar charts for channel comparison and a line chart for monthly trend. Audience: the founder. Tone: direct, numbers-first, no filler.

Why it works: it dictates slide count, section order and chart types — so you get editable chart objects where you want them instead of decorative art.

2 — The investor pitch deck

A 10-slide seed pitch for a UAE proptech that qualifies property buyers before agents call them. Standard order: problem, solution, product, market size for GCC real estate tech, business model, traction, competition, team, roadmap, ask. One idea per slide, minimal text, one strong visual each. Investor-grade minimalism.

Why it works: naming the canonical pitch order stops the model inventing a structure investors don't recognise.

3 — The market research deck (upload-driven)

Using the attached Q2 Dubai transaction report, build an 8-slide summary for clients: headline market movement, price per sqft by community, off-plan vs secondary split, buyer nationality mix, and what it means for buyers in the next quarter. Every figure must come from the attached document. Cite the page for each statistic.

Why it works: "every figure must come from the attached document" is the single best line for defeating hallucinated statistics.

4 — The creative campaign proposal (Visual mode)

A 9-slide campaign proposal for launching a waterfront residential project to European investors. Cover, market insight, the big idea, channel plan across Meta and Google, creative direction, media budget split, timeline, KPIs, next steps. Bold, editorial, dark palette with a single accent colour. Custom illustrations, no stock photography.

Why it works: campaign work is judged on the look, so this one earns Visual mode — and banning stock photography is what pushes it into Nano Banana Pro's custom graphics.

5 — The training deck

A 15-slide onboarding deck teaching new real estate agents how to qualify a lead: budget, financing, timeline, decision-maker. Include a flowchart of the qualification sequence and a slide of exact scripts for each question. Practical, no theory, designed to be taught from.

Why it works: asking explicitly for a flowchart gets an editable diagram object rather than a wall of bullets.

6 — The rescue job

Take the attached PowerPoint and rebuild it: same content and data, better structure, cleaner design, tighter copy. Cut anything repetitive, merge weak slides, keep every number exactly as it appears in the original.

Why it works: rebuilding beats generating when you already have content. "Keep every number exactly" is your safety rail.

Why Native Charts Actually Matter

Worth spelling out, because it's the reason to pick this tool at all.

Most AI presentation tools generate a chart by producing a picture of one. It looks fine in the preview. Then reality arrives: the CFO wants the figure updated, a client's actuals came in different, someone spots a typo in an axis label. With an image, none of that is editable — you rebuild the chart in Excel, screenshot it, and paste it back. Every single time.

A native chart object carries its data with it. In PowerPoint you right-click, open the data sheet, change the cell — and the chart redraws in the deck's styling. Kimi's own documentation is explicit that complex diagrams and charts are rendered as native components rather than static images, and that they stay adjustable after export.

Three places this pays off:

• Numbers move late. Business decks are edited hours before the meeting. Editable charts turn a rebuild into a keystroke.
• The deck outlives you. Hand it to a colleague and they can maintain it. An image-based deck dies the moment its author leaves.
• Google Slides and Keynote too. Export paths exist beyond PowerPoint, so the deck isn't trapped in one ecosystem.

Same principle I push on clients about owning your lead data rather than renting it: output you can edit is an asset, output you can only look at is a receipt.

Kimi Slides vs Gamma vs Canva vs PowerPoint

	Kimi Slides	Gamma	Canva	PowerPoint Copilot
Best at	Research depth + editable charts	Fastest first draft	Brand control and templates	Living inside Microsoft 365
Speed	5–10 min (Visual) / 30–60 min (Adaptive)	~30 seconds to a draft	Minutes	Minutes
.pptx export	Real .pptx, native chart objects	Common complaint: layouts and fonts shift	Solid PowerPoint compatibility	Native — it is PowerPoint
Custom branding	Template upload exists; not a full brand system	Themes, limited	Strongest of the four	Inherits your corporate template
Works from your docs	Strong — PDF, Word, Excel, PPTX, images	Basic	Basic	Strong within your tenant
Free tier	Real, credit-metered	Credit-based	Generous	Requires a paid Copilot licence

Where Kimi wins: research from your own documents, native editable charts, genuine .pptx. Where it loses: raw speed (Gamma), strict brand systems (Canva), and the free-tier queue.

My honest read: Gamma for something you'll share as a link. Canva when brand compliance is non-negotiable. Kimi when the deck carries data that someone will edit later. For most business decks that third case is the common one, and it's the one the others handle worst.

Honest Limits

• Verify every number. AI research can produce outdated or misattributed figures, stated with total confidence. Never let a generated statistic reach a client unchecked. Uploading your own source documents is the fix, not a promise from the model.
• Branding isn't fully solved. The feature page advertises uploading your company template, but independent reviewers consistently report it isn't a full corporate brand system — expect to apply strict brand styling after export. For agency-grade brand compliance, budget a cleanup pass.
• Free tier queues. Peak hours mean waiting. Combined with Adaptive's 30–60 minutes, this is not a fifteen-minutes-before-the-meeting tool.
• Very complex diagrams may stay images. Simple charts and SmartArt come through native; elaborate network diagrams sometimes don't.
• Availability is genuinely uncertain right now. With paid tiers paused, you can't buy your way past the free-tier limits at the moment. That's unusual, and it may have changed by the time you read this.
• It's moving fast. Modes, models and pricing all shifted in the last few months. Everything here was verified on 12 August 2026 — check the live site before making decisions on it.

Where This Fits

Same rule I applied to Design Arena and OpenHuman: test free, adopt narrow, expand only when the tool earns it. For Kimi Slides that means running one real deck — ideally a rescue job on a presentation you already have — and doing the double-click-the-chart test after export. If the chart opens its data grid, you've found a tool worth keeping. If you're building income around AI skills, "I can turn your messy report into an editable client deck in an afternoon" is a service people pay for today.

---

Found this useful? I break down AI tools like this — what's real, what's hype, and how to actually use them — on Instagram at @shaffay_bajwa. If you'd rather have the decks, content and campaigns built for you, that's what WIYO Marketing does — start a conversation whenever you're ready.