Kimi Slides - The Complete Guide

There's a deeply familiar frustration in the modern workflow: you know exactly what you want to say in your presentation, but you lose hours — sometimes a whole day — just saying it in slides. Choosing layouts. Compressing paragraphs to bullets. Finding visuals that don't look like 2012 clip art. Aligning text boxes. Rewriting everything when the stakeholder asks you to cut from 24 slides to 14. The problem isn't the ideas. It's the mechanical execution of packaging them.

AI presentation tools have proliferated as a response. Most of them solve part of the problem but create new frustrations: they generate content that sounds like a language model wrote it — generic headings, filler bullets, forgettable structure — and the visual quality is either stock-photo generic or aggressively inconsistent between slides. You still end up spending significant time fixing the output rather than delivering it.

Kimi Slides takes a structurally different approach, and it's worth understanding why. By combining Kimi's K2 Thinking model for research, argumentation, and slide copy with Google's Nano Banana Pro for custom per-slide image generation, it routes each type of work to the AI system most capable of handling it. The result is output that is simultaneously more substantive and more visually coherent than what single-model tools produce.

With the K2.6 update in April 2026, Kimi Slides gained meaningful new capabilities: upload your own visual references to define the exact style you want, editable SmartArt like timelines, flowcharts, and funnels, stronger source integration that can pull citations directly into your slides, and a more reliable generation pipeline overall. This guide covers the full picture — from how the tool works mechanically to how to get consistently good output from it.

What is Kimi Slides?

Kimi Slides is a specialized AI presentation agent built inside the Kimi+ ecosystem — Moonshot AI's suite of productivity tools that includes Kimi Docs, Kimi Sheets, Kimi Websites, and Kimi Code alongside Slides. It is not a standalone application; it runs as a dedicated presentation agent within the Kimi platform, sharing the same model infrastructure, context window, and tool-use capabilities as Kimi's core AI. You access it directly from kimi.com or the Kimi App — no separate install or account required.

At its foundation, Kimi Slides is a two-model system. The K2 Thinking model handles research, content structuring, logical argument organization, and slide copywriting — the parts that require genuine understanding of a topic and the skill to express ideas as clear, concise, impactful slides. Google's Nano Banana Pro (Gemini 3 Pro Image) handles custom image generation, infographic creation, illustration, and chart visual design — the parts that require producing visuals that actually match and reinforce the specific content on each slide.

This dual-model architecture is what separates Kimi Slides from tools that pass everything through a single language model. Language models are competent at reasoning and writing but genuinely poor at producing coherent visuals. Image generation models are powerful at visuals but have no understanding of slide narrative, content hierarchy, or logical argument structure. Kimi Slides routes each type of work to the model built for it.

Kimi Slides fits within Moonshot AI's broader "Office Pilot" direction — AI agents that handle the full lifecycle of common office deliverables rather than just assisting with individual steps. The positioning is deliberate: you describe what you need and receive a finished, presentable output. Slides, Docs, Sheets, and Websites are all part of this output-complete vision.

Adaptive vs Visual: Choosing the Right Mode

The most consequential decision in Kimi Slides is choosing your creation mode before you start. Adaptive and Visual are not simply "slow vs fast" variants — they produce fundamentally different types of output, draw on different AI systems at different depths, and are suited to different workflows and audiences.

Most professional users find that the mode choice maps naturally to the audience and purpose of the presentation. Adaptive mode produces output that analysts, consultants, and researchers need — content with integrity that withstands scrutiny. Visual mode produces output that wins pitches and commands attention on conference stages — decks that communicate "deliberate" rather than "AI-generated."

Nano Banana Pro: Why It Matters for Slides

Nano Banana Pro is Google's advanced image generation model — specifically the Gemini 3 Pro Image variant (the "Thinking" model option in Gemini's interface, distinct from the standard Nano Banana / Gemini 2.5 Flash Image). Within Kimi Slides, it is the engine behind all of Visual mode's custom illustration and infographic generation.

What makes Nano Banana Pro specifically valuable for presentation work - as opposed to other image models like DALL-E 3 or Midjourney — comes down to two capabilities that matter enormously in slides: text rendering quality and cross-slide visual coherence.

On text rendering: Nano Banana Pro generates legible, correctly spelled, appropriately formatted text within images far more reliably than competing image models. For slides that contain labeled charts, infographic statistics, annotated diagrams, or callout values, this is not a secondary consideration — it is the difference between a slide that works as-is and one that requires manual correction before it is presentable. DALL-E and Midjourney regularly produce hallucinated or misspelled text in images; Nano Banana Pro gets it right significantly more consistently.

On visual coherence: when you generate a 20-slide deck through Kimi Slides' Visual mode, each slide's imagery follows the same design direction — consistent color palette, consistent illustration style, consistent compositional rules. The visuals look like they came from the same creative brief. This cohesion is what separates a professional-grade deck from an AI-generated one where each slide's visual was clearly generated independently.

Chart Parsing: The Most Underrated Feature

Nano Banana Pro's most technically impressive capability in Kimi Slides is chart parsing. When you upload a document or image containing charts or graphs, Nano Banana Pro autonomously:

• Extracts the textual and numerical data from the chart image
• Strips away the static original labels and visual encoding
• Rebuilds the chart as a native PowerPoint chart object - not a screenshot or rasterized image locked in place
• Overlays clean, editable text boxes on the rebuilt chart with accurate labels
• Preserves all underlying data so you can update values and labels in PowerPoint after export

This capability solves one of the most tedious recurrent problems in professional presentation work: the slide that must contain a chart from a PDF report that needs to be updated, reformatted, or rebranded. What previously required manually recreating the chart in PowerPoint from the original data source is now a single upload operation. The rebuilt chart object is fully editable in your PPTX export.

Agentic Slides: Converting Documents into Decks

Agentic Slides is Kimi's document-to-presentation conversion capability - arguably the most practical feature for professionals who regularly need to transform existing work products into slide decks. Upload your source material and Kimi's agent autonomously analyzes its structure, identifies the key points worth converting to slides, makes editorial decisions about content hierarchy and emphasis, and produces a presentation that reflects the logic of the source document rather than just reproducing its text in a different visual container.

This is genuinely different from mechanical text extraction. When you upload a 40-page research report, Kimi doesn't divide it into equal text chunks and format them as slides. It reads the argument structure, identifies the core findings, decides which supporting evidence deserves its own slide versus which should appear as a data point, and constructs a narrative arc appropriate for a presentation rather than a document. The output is a first-draft deck that a skilled analyst might have produced from the same source material — with the hours of manual extraction work compressed into minutes.

Supported Input Formats

• PDFs - research papers, reports, proposals, contracts, books
• Word documents (.docx) - briefs, drafts, notes, structured outlines
• Excel spreadsheets (.xlsx) - data files, financial models, analysis outputs
• Image files — charts, diagrams, screenshots, visual references
• Raw text / paste content - rough notes, URLs, structured outlines
• Up to 50 files simultaneously - batch document analysis for comprehensive decks

K2 Live Search Integration

In Adaptive mode, Kimi Slides doesn't limit itself to the content you provide - it integrates an agentic K2 search tool that pulls current data, statistics, and contextual information from the web during generation. This is the same research capability that powers Kimi's Deep Research feature, applied specifically to enriching presentation content with fresh, relevant material.

In practice: if you ask Kimi to create a presentation on renewable energy investment and provide a brief outline, the resulting deck will contain actual 2025–2026 investment figures, current project examples, and recent policy developments sourced autonomously during generation. The K2 search enriches presentations beyond what training data alone would support, particularly for topics where recency matters - market analysis, technology updates, policy landscapes, scientific developments.

What K2.6 Brought to Kimi Slides

When Kimi K2.6 launched in April 2026, it brought meaningful improvements to the Slides feature that go beyond incremental refinement. The upgrades target three areas that had been limiting factors in the previous version: visual customization, structural content types, and source credibility.

Key Features of Kimi Slides

How to Use Kimi Slides: Step by Step

Who Kimi Slides Is For: Real Use Cases

Kimi Slides Pricing Across Membership Tiers

Kimi Slides is included in all Kimi membership tiers. Under Kimi's unified credit system, Slides usage draws from the same shared credit pool as other Kimi+ features — so you can allocate your monthly credits toward Slides, Deep Research, Agent tasks, or any other feature based on your actual workflow each month.

Kimi Slides vs Other AI Presentation Tools

Comparison as of April 2026. Features and pricing change frequently - verify on official pages before committing.

Tips for Getting the Best Results

• Never skip the outline review. The editable outline is your single highest-leverage intervention point. The AI is excellent at filling in content within a solid structure; the structure itself is where human judgment adds the most value. Spending 5 minutes refining the outline saves 30 minutes of post-generation fixes.
• Use narrative headings, not topic labels. "Market Overview" is a topic label. "The market is growing faster than incumbents can respond" is a narrative heading. Narrative headings tell a story and produce more compelling slide copy from the AI. This change alone significantly improves output quality.
• Use visual references for design-critical presentations. K2.6's new visual reference upload feature is the most impactful way to control output style. Upload 2–3 sample slides, a style reference, or even a screenshot of the aesthetic you want, and Kimi will match it far more accurately than trying to describe the style in text.
• Constrain slide count strategically. Start with 8–10 slides for initial generation. Review structure and quality, then expand in subsequent requests. Fewer slides initially means faster iteration cycles if the approach needs adjustment. You can always ask Kimi to add a section afterward.
• For document conversion, summarize emphasis first. Before uploading a long document, add a brief note about which sections matter most and what the key message should be. This reduces the chance that the AI emphasizes supporting details over main findings.
• Do all text editing in-browser, then export for visuals. Visual elements in Nano Banana Pro slides require full regeneration to change within Kimi. Edit all copy in the browser, then export to PPTX for any visual element adjustments in PowerPoint. Don't fight the workflow - use each tool for what it does well.
• For urgent presentations, use Visual mode over Adaptive. When time is the constraint, Visual mode's 5-10 minute generation is significantly more practical. The design quality is high enough that a Visual mode deck, with minimal text refinement, can often be presented directly. Adaptive mode's depth requires the investment of 30-60 minutes and benefits from a subsequent review cycle.

Honest Limitations of Kimi Slides

Visual elements are locked post-generation

The most significant limitation in the current version: in Kimi's browser interface, only text is editable after Nano Banana Pro generates a Visual mode slide. Icons, illustrations, infographic elements, and non-text visuals are locked — modifying them requires regenerating the slide from scratch, which costs time and credit. For presentations where you anticipate multiple rounds of visual iteration, plan your workflow around this: make visual decisions at the prompt stage, not after generation. Export to PPTX for any fine-grained visual control.

No custom corporate brand templates

Kimi Slides does not yet support uploading your organization's PowerPoint template — with your brand fonts, color schemes, logo placements, and master slide layouts — and having Kimi generate slides that conform to it automatically. The visual reference upload feature (K2.6) is a partial workaround, but it does not provide the same degree of brand fidelity as template-based tools like Beautiful.ai. This is a material limitation for enterprise users with strict brand guidelines, and reportedly on the product roadmap.

Adaptive mode is genuinely slow

The 30–60 minute Adaptive mode runtime is accurate and reflects the genuine depth of research Kimi conducts. For users accustomed to faster tools, this is a real friction point — particularly if the initial structure needs revision and a second generation pass. The wait is justified by output quality for the right use cases, but Adaptive mode is not appropriate for last-minute or truly time-pressured work.

Queue delays on free tier during peak hours

Free tier (Adagio) users experience generation queue delays during peak usage periods — typically evening Beijing time (approximately 6-10 PM CST). For time-sensitive work on the free tier, schedule generation outside these windows or upgrade to Moderato for priority queue access, which significantly reduces waiting on Visual mode generation in particular.

Verify facts from Adaptive mode before presenting professionally

Despite K2's live web research capability, Adaptive mode can produce inaccurate statistics, outdated figures, or misattributed claims - particularly on niche topics with thin or inconsistent web coverage. For any presentation that includes specific numerical claims or citations, verify against primary sources before presenting in a professional context. The research assistance dramatically reduces manual effort; it does not eliminate verification responsibility.

The Bottom Line: Is Kimi Slides Worth It?

For most professional presentation creation workflows in 2026, Kimi Slides is the most technically capable AI presentation tool available - and the strongest argument for that claim is the combination of K2 Thinking's research depth with Nano Banana Pro's visual quality that no other tool currently matches at any price tier. The K2.6 upgrade added SmartArt, visual reference uploads, and source integration, making the tool meaningfully more capable than its previous iteration.

The free Adagio tier is genuinely useful. Adaptive mode with document upload and PPTX export is a real capability that covers the core use case - converting a topic or document into a structured, presentable slide deck - without any subscription cost. For anyone who regularly creates presentations and hasn't tried Kimi Slides, the free tier alone justifies a test run today.

The Moderato upgrade at $19/mo is easy to justify if you create presentations with any regularity. Visual mode with Nano Banana Pro produces output that is visually closer to commissioned designer work than what any competing AI tool produces at this price point. The 5-10 minute generation time is fast enough for practical use, and the chart-to-native-PPTX-object parsing eliminates one of the most tedious recurring tasks in professional deck work.

The limitations are real - locked visual elements, no custom brand templates yet, slow Adaptive mode, queue delays on the free tier - but none are dealbreakers for the tool's core use cases. They define the edges of where Kimi Slides fits natively versus where it needs supplementation from PowerPoint post-export. Treat it as a high-quality first-draft engine with excellent structural intelligence and strong visual defaults, not as a complete replacement for the final polish that every professional presentation requires.

Related Articles