explainx.ainewsletter3.5k
TrendingNewsPathwaysSkills
Pricing
explainx.ai

Upskill in AI — 16 free pathways, live workshops & bootcamps, and 50+ courses from practitioners. Plus the skills, tools, and MCP servers to practice on.

follow us

corporate training

support@explainx.ai

get started

Find your pathTake Free Evaluation

learn

pathways — start freeworkshopsbootcampscoursescertificationsmock testsexplainx universitycorporate traininglearn skills & mcp

discover

skillsmcp serversexplainx mcptoolsagentsllmsdesignsdictionaryagi trackerranks

company

aboutvisionmissionteaminstructorscommunityhackathonscareers

content

daily AI newsstate of AI — live resultsblogreleasespromptsgeneratorsresource libraryfor LLMsexplainx.ai kids

solutions

all solutionsdeveloper upskillingmarketing upskillingproduct manager upskillingleadership upskilling

newsletter · weekly

Get AI news, tools, and insights in your inbox.

supportprivacytermsdata rightshow we create contentsubmission guidelines

© 2026 AISOLO Technologies Pvt Ltd

On this page

  • TL;DR
  • What the walkthrough actually shows
  • Where it sits in Moonshot’s stack
  • How to actually use it (practitioner path)
  • Kimi Slides vs the rest of the 2026 slide stack
  • What people are asking
  • Honest limitations
  • Closing
  • Related on explainx.ai
← Back to blog

explainx / blog

Kimi Slides: Research-to-PPTX That Stays Editable

Moonshot’s Kimi Slides turns a topic into a researched, branded deck with native charts and SmartArt — then exports editable PPTX, not flattened images.

Aug 13, 2026·8 min read·Yash Thakker
Kimi SlidesMoonshot AIKimi K3PresentationsKnowledge Work
go deep
Kimi Slides: Research-to-PPTX That Stays Editable

Most AI slide tools give you a pretty PDF you cannot fix. Kimi Slides is Moonshot AI’s bet that the file you keep should still be a PowerPoint.

In the official walkthrough (YouTube), presenter Rachel walks a full loop: give Kimi a topic, let it research and structure a consulting-style story, keep charts and SmartArt as native objects, comment on a slide to iterate, roll back via version history, rebuild a technical diagram from an image or PDF, then export to PowerPoint. Posters, infographics, and social cards are in the same pipeline.

That is the office-work layer sitting on Kimi K3 and Kimi Work — not a new foundation model, a deliverable.

Official Kimi Slides walkthrough: research, native charts/SmartArt, comments, version history, PPTX export.

TL;DR

table · 2 cols
QuestionDirect answer
What is it?AI deck generator inside Kimi Work, powered by Kimi K3
Killer feature?Native editable charts / SmartArt in exported PPTX
Free?Adaptive mode on Adagio (free); Visual mode from $19/mo
How long?Visual ~5–10 min; Adaptive ~30–60 min (estimates, not SLAs)
InputsTopic, text, PDF/Word/Excel/Markdown/images, your PPTX template
ExportsPowerPoint, Google Slides, PNG, or present in-browser
Trust the citations?No — Moonshot does not guarantee fact accuracy
Best skip?Strict brand manuals, confidential PII uploads to a Beijing-based vendor
Weekly digest3.5k readers

Catch up on AI

Curated AI updates on agents, skills, and MCP — delivered to your inbox. Unsubscribe anytime.

What the walkthrough actually shows

Stick to the video, not the landing-page poetry.

1. Research before layout. A presentable deck needs structure, cohesive design, and editing rounds. Kimi takes the whole pipeline: find data, organize a focused story, then format. That is the opposite of “pick a template and fill bullets.”

2. Native objects, not screenshots. Charts from business data, project milestones on a timeline, relationships as SmartArt. You edit the data behind charts, move shapes, and adjust SmartArt nodes and connectors. This is the feature Gamma-style flattened exports usually lose.

3. Brand from a reference file. Start from a Kimi template or upload a PowerPoint, image, or PDF. Kimi pulls colors and layouts so the new deck matches the reference — with the official caveat that template support is directional, not pixel-perfect.

4. Talk to the deck. First drafts are rarely final. Comment on a specific slide; Kimi jumps in. Version history records every draft so you can roll back.

5. Diagram rebuild. Upload an image or PDF of a complex technical diagram. Kimi claims it recognizes text, shapes, connectors, and formulas and recreates them as editable PowerPoint elements.

6. Ship in Office. Export to PowerPoint and keep collaborating. Same generator also does posters, infographics, and social cards — still editable.

Official product page: kimi.com/features/slides.

Where it sits in Moonshot’s stack

table · 2 cols
LayerJob
Kimi K32.8T MoE, 1M-token context — research + structure
Kimi WorkDesktop agent: local files, WebBridge, swarm → PPT/Excel
Kimi SlidesThe slide surface: outline → design → editable export

Kimi Work already advertised swarm output into PowerPoint. Slides is the interactive editor on top of that: comments, version history, brand references, diagram OCR. Moonshot’s Japan account framed a Kimi Work tutorial series around Slides on August 3, 2026 — this is a promoted feature, not necessarily a brand-new launch day.

Two generation modes (from Kimi’s docs; times are estimates):

table · 3 cols
AdaptiveVisual
PriorityContent + researchVisual impact
EngineK3 layoutsNano Banana Pro illustrations
Typical time30–60 minutes5–10 minutes
PlanAdagio (free)Moderato ($19/mo) and up
UseReports, technical decksPitches, keynotes

Support pages disagree on speed (“a few minutes” vs 30–60). Plan for the high number if you have a hard meeting.

How to actually use it (practitioner path)

  1. Dump source, not vibes. Upload the PDFs, spreadsheet, and last quarter’s deck. Caps cited in third-party write-ups: 100MB per file, 50 files per upload — verify on the live help center.
  2. Write the decision, not the topic. “Q3 board: approve EU data-center capex; every figure with year + source; close with three risk scenarios” beats “data center market.”
  3. Review the outline before render. Kimi tends to give every section equal slide count. Kill bloat here — regeneration after layout is expensive in Adaptive.
  4. Check charts and citations by hand. Traceable on-slide citations speed audit; they do not replace it. Moonshot’s own docs disclaim accuracy.
  5. Export PPTX, then do the last 10%. Fonts missing locally, cropped images, broken links. Same last-mile as Copilot drafts.
  6. Use comments + version history instead of regenerating the whole deck for one number.

If you already run Kimi Work Cron + finance data, a recurring deck (weekly KPI, monthly board) is the high-leverage loop — same pattern as Claude Cowork scheduled jobs, different vendor.

Kimi Slides vs the rest of the 2026 slide stack

Name the category honestly. Do not treat this as a feature-complete bake-off.

table · 2 cols
NeedReach for
Research-heavy first draft, native PPTX objects, free AdaptiveKimi Slides
Fast pretty web deck, design templatesGamma / Beautiful.ai (design-first SaaS)
Stay inside Microsoft 365 + existing slide masterCopilot in PowerPoint
Open-source AI office (Docs/Sheets/Slides Alpha)Genspark GenOffice
Knowledge-work prompting, not a deck factoryClaude for Work workshop / Claude vs ChatGPT for work
Agent design systems for UI, not slidesDESIGN.md templates

Kimi’s wedge: research-before-design + objects you can still edit in Office. Kimi’s gap: brand fidelity and template variety vs dedicated design tools; Adaptive is slow; citations are unaudited.

What people are asking

Is this just Gamma with K3?

No. Gamma wins on speed and visual taste for many teams. Kimi wins if you need a spreadsheet-backed chart you can fix in PowerPoint and a research pass that cites sources on-slide. If you only need a 12-slide pitch by lunch, Visual mode (paid) is the closer analogue — still not Gamma’s template library.

Why would Adaptive take an hour?

K3 is doing retrieval, outline, and layout in one job. That is closer to an agent swarm than a “make slides” button. Use Visual when the story already exists in your uploads; use Adaptive when the model must find the story.

Can I trust it for a client deck?

As a first draft, yes if you verify every number. As a send-as-is artifact, no. Same rule as any LLM research product. Editable PPTX makes correction cheap; it does not make the model honest.

Should EU/UK teams upload customer data?

Moonshot is Beijing-headquartered. There is no EU/UK adequacy decision for China. Public-data decks are a different risk from customer files, health data, or unreleased financials. Treat uploads as a transfer and check counsel before you paste a CRM export.

Does the diagram-rebuild actually work?

The walkthrough claims OCR of text, shapes, connectors, and formulas into editable elements. Treat that as a demo claim until you try it on your architecture diagrams. Expect cleanup on dense Visio/PDF exports.

Honest limitations

  • Walkthrough is marketing; we did not independently time Adaptive vs Visual on a production deck.
  • Citation accuracy is explicitly not guaranteed.
  • Template matching is style-direction, not a brand-manual compiler.
  • Timing estimates conflict across Kimi’s own pages.
  • Older Office suites: variable PPTX compatibility.
  • Free-plan queues can stretch Adaptive further.
  • Not a substitute for Kimi Work WebBridge/swarm if your bottleneck is collecting files from the live web.

Closing

Kimi Slides is Moonshot packaging K3 as a file you can present. The walkthrough’s useful idea is not “AI makes pretty slides.” It is research → native Office objects → comments and version history, so the last 10% stays in PowerPoint instead of in a locked PDF.

Try Adaptive on a public-data internal briefing first. Keep confidential decks in Copilot or a local GenOffice path until you have a data-transfer story.

Follow @explainx_ai when Moonshot ships the rest of the Kimi Work tutorial series.

Related on explainx.ai

  • Kimi Work — desktop agent, WebBridge, swarm to PPT/Excel
  • Kimi K3 open weights — 2.8T, 1M context
  • Kimi K3 API / pricing primer
  • Genspark GenOffice — open-source AI office
  • Claude vs ChatGPT for work
  • Top 10 Claude Cowork use cases
  • Claude for Work live workshop
  • DESIGN.md for AI-generated UI

Sources

  • Kimi Slides walkthrough (YouTube) — research, native charts/SmartArt, comments, version history, diagram rebuild, PPTX export
  • Kimi Slides product page
  • Kimi Work
  • Moonshot Japan / Kimi Work tutorial series (Aug 3, 2026) — Slides as tutorial #1

Features, plan names, and timing estimates reflect Moonshot’s public walkthrough and Kimi docs as of August 13, 2026. Quotas, Visual-mode pricing, and Adaptive duration can change — verify on kimi.com before a client deadline. Not affiliated with Moonshot AI.

Spotted something out of date? Let us know.
Yash Thakker

Written by

Yash Thakker

Yash is an AI expert with over 300K learners. Join his workshops →

Related posts

Aug 8, 2026

Kimi K3 "Escaped Containment"? We Could Not Verify the Claim

explainx.ai previously reported that Moonshot AI's Kimi K3 reached the open internet during a security test. A source audit could not locate the cited WIRED story or any first-party incident disclosure, so this page now records the correction, the checks performed, and the facts that remain verified.

Jul 27, 2026

Kimi K3 Open Weights Are Live — 2.8T Parameters, Day-0 on Together and Modal

Moonshot AI published open-source weights for Kimi K3 on July 26, 2026 — roughly a day ahead of its own July 27 target — putting a 2.8-trillion-parameter, 1M-context frontier model on Hugging Face for free download. Together AI and Modal both announced day-0 hosted access. Here's what's confirmed, what's still a claim, and how the release lands amid a live US policy fight over open-weight Chinese models.

Jul 21, 2026

Kimi Work: Moonshot Desktop Agent — WebBridge, Swarm, and Cron Guide

Moonshot's Kimi Work mounts your folders, drives Chrome/Edge via WebBridge, schedules Cron agent runs, and coordinates up to 300 K3 sub-agents into PowerPoint and Excel. explainx.ai maps who it fits, honest platform limits (Apple Silicon Mac only, sleep stops tasks), and how it compares to web Kimi, ChatGPT desktop, Claude Cowork, OpenClaw, and jcode.