Convas
Start creating

Changelog

Gemini 4 Argon, plus exports that don’t need you to wait

Gemini 4 Argon is in the model list—great when you send references or need long back-and-forth rewrites. PDF, PPTX, and image ZIP exports from chat also keep running in the background so you can step away and download when they’re ready.

Reach for Gemini 4 Argon when the work gets heavy

Google’s Gemini 4 Argon is now in the chat model picker and project settings. It’s built for multi-step work—slides, brief, and outline staying in sync, plus reference images to say “match this vibe.” For quick daily tweaks, Gemini 3.8 Flash and other lighter options are still there.

Gemini 4 Argon in the composer model picker
Switch to Gemini 4 Argon from the composer or project settings.

Exports no longer need a babysitter

Like batch slide rendering, when you ask the agent in chat for a PDF, presentation PPTX, or a zip of slide images, packing runs in the cloud. Chat can wrap up first; progress and download stay on the tool card, and the top-bar download matches what the agent exported.

  • PDF due minutes before a meeting, but you still want one more copy pass
  • Full-deck PPTX takes a while—you’d rather not watch the spinner
  • Rendering or exporting a long deck while you step away and collect files later
Great when you need to↓
  • Research or proposal decks where Google’s top tier should steer many rounds of edits
  • Slides and PDF due together without being glued to the browser
  • You already rely on background rendering and want exports to feel the same

Changelog

Draft, edit, and present—without leaving Convas.

Agent-written scripts live in project materials; open the notes drawer while presenting to edit on the fly. Presenter view puts the audience on fullscreen while a second window shows notes, next slide, and pacing. Share links hide notes by default—turn on “Visitors can view speaker notes” when you want them on the guest page.

Where notes live—and how to edit them

Ask the Agent in chat to draft or refine per-slide scripts; they land in your project speaker-notes file, previewable under Brief › Project materials › Speaker notes. While presenting from the canvas, open the notes drawer from the toolbar—text follows the active slide and saves back to the project. Editable PowerPoint export carries the same notes into slide remarks.

Canvas notes drawer under the slide preview with the current page script
Notes drawer on canvas: edit per slide; changes save to your project file.

Two screens while you present

Turn on presenter mode in presentation settings, go fullscreen, then open presenter view in a second window. The audience screen stays on slides; the presenter window stays in sync with thumbnails for current and next slide, page notes, timers, and pace hints. If the browser blocks the popup, use the on-screen button to open it.

Presenter view with current and next thumbnails, notes, and timers
Presenter view: notes, next slide, and pacing on your screen; audience stays on fullscreen slides.

Share links and notes

Read-only share links still default to slides only—guests cannot see your workspace or your script. For training or async review, enable “Visitors can view speaker notes” in the share popover before copying the link; guests toggle a notes drawer that follows the slide they are on. Turn the switch off or revoke the link anytime.

Great when you need to↓
  • Rehearse a pitch: notes and next slide on your screen, deck on the projector
  • Dual-monitor talks: browser fullscreen for the room, pacing tools on yours
  • Client links without scripts; internal links with opt-in speaker notes

Changelog

Data slides and research decks—two new asset types in your library

Everyday KPI charts stay editable on the canvas; heatmaps, correlations, and other heavy stats can land on slides too. Embedded figures from PDFs extract in bulk—no manual cropping—preview under Brief › Project materials › Analysis or Embedded.

Two ways to chart your spreadsheet

Upload Excel or CSV and everyday KPI bars, trend lines, and share charts work as before—numbers come from your file, and you can still tweak values on the canvas refine panel. When a slide needs heatmaps, regression trends, or multi-factor correlations that pivot tables cannot express, describe the analysis in chat; the Agent runs Python on your spreadsheet to compute the chart—not guessed KPIs—and places it on your slide. These ship as finished figures—update the spreadsheet and ask for a refresh if the numbers change.

Figures from PDFs, no manual cropping

For thesis, research, or industry decks, besides text summaries you can batch-extract embedded figures from PDFs and cite them on slides. That reuses the paper’s originals—it does not digitize curve data from chart screenshots. When supplementary spreadsheets exist, Excel/CSV still wins for editable charts.

Find them in the Brief tab

  • Brief › Project materials › Analysis: stats charts from your tables—heatmaps, regression, and more
  • Brief › Project materials › Embedded: figures cut from PDFs, ready for slides
  • Sources / Illustrations / Research: originals under Sources; illustrations and research notes in the Project materials tree
Great when you need to↓
  • Weekly ops or finance decks from Excel, with last-minute KPI tweaks on canvas
  • Insight slides that need correlation, distribution, or regression—not just bars and lines
  • Academic or research decks that must cite original PDF figures without manual cropping

Changelog

SVG generation, editing & vectorization

The Agent can generate decorative vectors, turn bitmaps into SVG, tweak existing SVGs, or add motion—saved to your project library. Pick Vector or Vector · Fine in Settings › Preferences; chat and batch render stream live previews.

Four SVG capabilities—from generate to motion

Beyond structural diagrams drawn at render time, Convas supports standalone SVG assets: in chat or batch render, the Agent can generate flat icons and decorative art, vectorize PNG/JPG already in the project, tweak existing SVGs (recolor, copy edits, small shape changes), or add motion (logo reveals, loading loops, and more). Files land in your project library with streaming SVG draft previews.

Four capabilities

  • Generate: describe decorative vector art and get a scalable SVG file
  • Vectorize: turn an existing bitmap into SVG—best for simple icons and flat graphics
  • Edit: recolor, update copy, or make small shape tweaks on an existing SVG
  • Animate: add motion to an existing SVG and output animated SVG

Choosing an engine

  • Vector (default): generate, vectorize, edit, animate, and live preview—billed by usage in Credits
  • Vector · Fine: better for complex briefs and refined motion
  • Gemini / DeepSeek (experimental): generate and vectorize only, billed similarly to Chat

Flowcharts, architecture sketches, and timelines are still drawn directly at render time—not through this path. See the Visual elements doc for details.

Great when you need to↓
  • Build a set of flat icons or decorative art that stays crisp when scaled
  • Turn an existing logo or simple image into SVG before placing it on a slide
  • Recolor generated icons to match brand tokens or add a logo reveal animation
  • Generate or refine SVG assets inside the same batch while rendering slides

Changelog

DeepSeek V4.1 Flash: images in chat by default

The default chat model is now DeepSeek V4.1 Flash—native multimodal, so references and screenshots go straight to the model. Stronger reasoning, faster responses, lower cost. DeepSeek is a single choice in the model selector.

One DeepSeek, native vision

Convas now defaults to DeepSeek V4.1 Flash. Images no longer take a side path—drop in references, slide previews, or charts from your materials and the model reads them natively while it keeps helping with brief, design, and slides. The model selector shows a single DeepSeek option—no more picking among Pro, Flash, or vision experiment tiers.

Same workflows, connected

  • Attach brand refs, competitor shots, or mockups—“make it feel like this”
  • Send a slide preview and ask “is the title too small?” or “is this palette too loud?”
  • Share charts or layouts from your materials so the AI grasps the visual direction you want
Great when you need to↓
  • Live in DeepSeek for outlines and reasoning but still need reference images in chat
  • Mark up preview screenshots before a meeting without switching models
  • Want the default model to stay strong, affordable, and image-ready

Changelog

Batch renders keep going after chat

When generating many slides from an outline, chat may finish quickly while rendering continues on the server—reopen the project to see thumbnail progress; the composer shows background rendering or queued status.

Rendering no longer tied to the chat turn

Batch slide rendering used to require keeping the browser tab open. When generating many pages from an outline, render jobs now queue and run on the server—chat can finish first while you edit the brief, tweak design, or step away.

Pick up where you left off

Reopen the project and the canvas thumbnail rail resumes live progress; the composer shows background rendering or queued status. The current chat turn—streaming reply, parsing documents, drafting an outline—still needs this tab open; closing or refreshing stops that turn, unlike batch rendering.

Great when you need to↓
  • Long decks take minutes—you do not want to watch the chat window the whole time
  • Keep editing brief or outline while slides render
  • Switch tabs briefly and come back to more finished thumbnails

Changelog

Pick real images from the web for your slides

When the assistant researches the web, it can surface images from news, product pages, or reports—you pick the ones you want in chat, they save into project assets, then land on slides. No hotlinked URLs. Atmosphere art still uses AI image generation.

When the pick card shows up

When you need real-world visuals—press photos, brand logos, product shots, charts from reports—the assistant may show a “Pick images” card after a web search or page fetch. Decorative backgrounds and mood illustrations stay on the AI image path, not this flow.

What you do

  • Review thumbnails and source hints; select 1–4 images (often 1 per slide)
  • Tap Use selected—images save into your project library, then go on slides
  • Tap Skip if nothing fits; the assistant can fall back to generated art or a placeholder instead of forcing a bad match

Versus AI-generated images

Web pick is for “must be real”—citable news frames, official product shots, report screenshots. AI generation is for “doesn’t need a source”—mood art, abstract backgrounds, stylized filler. Files you upload live in the brief library alongside both paths.

Great when you need to↓
  • Cite press or research imagery in a sector deck or fundraising story
  • Use real logos and product photos—no fabricated brand assets
  • Match visuals to a target market’s sources in multilingual decks
  • Stay in control of rights-sensitive images instead of silent hotlinking

Changelog

Drop a short video reference in chat

Product demos, screen recordings, reference clips—attach mp4 in chat with a video-capable model and the assistant reads visuals and narration to help with brief, outline, and slide copy.

Show the motion, not just tell it

Some asks don’t land in a screenshot—how the product moves, which beat in a screen recording matters, what the voice-over in a reference clip is stressing. Drag or pick a short mp4 in the chat box and the assistant uses picture and sound together to keep helping with brief, outline, or slide copy.

Same flow as images—pick a video-capable model

Choose a model that supports video in the selector to attach a clip in chat. If you switch to a model that doesn’t, remove the video before sending. One video per message and roughly up to 90 seconds works best for quick reference.

  • Attach a product demo—“match this motion and pacing”
  • Send a screen recording—“borrow this screen’s information architecture”
  • Share a reference clip so the assistant can pull voice-over points into brief or speaker notes
Great when you need to↓
  • UI motion or walkthroughs are hard to describe in text alone
  • The reference is a video—competitor reel, venue sample—and screenshots feel lossy
  • You already have mp4 in the brief library; @mention or upload works as reference in chat

Changelog

Fix formulas on the canvas

On code-mode slides, click a formula to edit LaTeX in the refine panel—block and inline math both work. Fix a symbol or switch display mode before the meeting without regenerating the slide in chat.

Click the formula, edit the LaTeX

Select a block or inline formula on the canvas and open the Formula section in refine: edit LaTeX directly, watch preview update live, then Apply to save. Integrals, fractions, matrices, and more render through KaTeX—PDF and high-fidelity PowerPoint export match what you see in preview.

Chat for new math, canvas for tweaks

Keep using chat when you need a full derivation or a new equation on the slide. When one subscript is wrong or you want inline vs block display, fixing it on the canvas is faster than regenerating the page. Charts, visuals, and formulas can be pinned on the same slide without getting in each other’s way.

Great when you need to↓
  • Tech or finance decks with lots of math—a symbol needs fixing before the meeting
  • Inline math mixed with body copy—you want to tweak LaTeX without rerunning the Agent
  • Chat already got the equation mostly right—you just need a final LaTeX pass

Changelog

Edit image slides where you click

When a slide is a finished image, you shouldn’t redo the whole page to fix one headline or swap one photo. Click the spot you want, say what to change, and leave the rest as is.

Feels like editing any other slide

Image-mode decks used to feel locked in once rendered—polished, but hard to tweak. Now click a headline, photo, or stat on the canvas, describe the change in plain language, and only that region updates. Preview refreshes right away.

Big changes in chat, small fixes on the canvas

Keep using chat when you want a new story, a new look, or a full re-layout. For last-mile polish—copy, assets, a number that’s off—click-to-edit is faster and avoids regenerating the entire slide.

Great when you need to↓
  • Swap a client logo or photo without rebuilding the layout
  • Fix a headline or figure without touching the rest of the slide
  • Patch small issues right before you share or present

Changelog

Pick the right stage ratio for your deck

Choose an aspect ratio when you create a project, and the whole deck stays in that shape—from generation through export. Widescreen, vertical, square—no cropping after the fact.

Match the ratio to the room

Widescreen 16:9 is no longer the only default. Pick 4:3 or 21:9 for the boardroom, 9:16 or 3:4 for short video, or 1:1 for social covers. Canvas, thumbnails, and exports all stay consistent—your stage is right before you present.

One ratio end to end

Set the format when you create the project; chat generation, canvas preview, downloads, and share links all honor it. On the homepage demo you can switch formats too, to preview how decks feel on different screens.

Common picks↓
  • Pitch or board meeting: 16:9 or 21:9 widescreen
  • Short-form social: 3:4 or 9:16 vertical
  • Covers and square posts: 1:1

Changelog

Fix chart numbers on the canvas

On code-mode slides, click a chart to edit values and axis labels—including charts built from uploaded spreadsheets. Update a figure before the meeting without reopening Excel or regenerating the slide.

Click the chart, edit the numbers

Select a chart on the canvas and adjust series values, axis labels, and key copy in the refine panel. Preview updates right away—ideal for data slides and trend charts before you walk into the room.

Chat for structure, canvas for digits

Keep using chat when you need a new chart type, a new story, or a full re-layout. When one number is off or a label has a typo, fixing it on the canvas is faster than regenerating the slide.

Great when you need to↓
  • Update a KPI minutes before a client meeting
  • Fix an axis label or legend without a full rerun
  • The data is mostly right—just one or two tweaks left

Changelog

Canvas tweaks: shift the whole deck or one line

You don’t need chat for every polish. Slide the canvas rail to shift colors and type across the deck; click a headline or paragraph to fix copy, color, and alignment in place—small edits apply right away.

Tune the whole deck

Open Tweaks from the canvas toolbar on code-mode slides and adjust primary colors, backgrounds, body size, spacing, and corner radius in the right rail—the whole deck shifts together. Great for “a bit darker” or “slightly larger type” before you lock the look.

Fix one block yourself

In edit mode, click a headline or paragraph and switch to refine: edit copy, color, and alignment directly—no chat message, no extra model turn. For chart numbers, use chart refine; for big rewrites, keep using chat.

Great when you need to↓
  • Feedback says “make it more corporate”—nudge colors and type
  • One typo in a title—fix and share
  • Alignment or color feels off—a quick slider beats a long prompt

Changelog

Upload a spreadsheet, charts you can trust

Drop Excel or CSV into your brief and build data slides from real calculations—not guessed KPIs. Once charts are on the slide, you can still adjust numbers on the canvas.

Spreadsheet in, chart out

Fundraising metrics, annual figures, survey rollups—upload the spreadsheet as project material first. Describe the trend or comparison you need in chat; the system reads structured data, aggregates, and builds charts on your slides. Numbers stay traceable so you can stand behind them in the room.

Adjust after the math

Auto-generated charts are usually ready to present. If one figure changes or an axis label needs clearer wording, click the chart on the canvas to tweak—no trip back to Excel, no full-slide rerun.

Great when you need to↓
  • Turn several tables into trend slides for a pitch or weekly report
  • Figures must match finance or ops—not “plausible” guesses
  • Ship a first chart draft, then fix one or two KPIs before the meeting

Changelog

Export to PowerPoint: editable or locked layout

For code-mode decks, pick at download time: an editable PowerPoint you can keep tweaking, or a locked full-slide-image version that matches the preview. PDF and per-page PNG ZIP live in the same menu.

Still editing in PowerPoint — go editable

When teammates need to tweak copy or chart data in PowerPoint after the meeting, export the editable version. Text and common charts stay as native objects when possible.

Layout locked — go high-fidelity

When the deck is final and you want to prevent accidental edits, export full-slide-image PPTX that matches the canvas. Image-mode decks take this path by default.

Same download menu↓
  • PDF: print, email attachments, read-only handoffs
  • Image ZIP: one PNG per slide for docs or archiving
  • Share links stay in the header—separate from download

Changelog

Ten slides, not ten waits

Once the outline is set, many slides can generate at once. The side tray shows progress per page—you can review finished slides while others are still rendering.

The whole deck moves at once

Cover, section, and content slides can advance together. Long decks no longer mean serial waiting—you can open finished pages on the canvas early and catch issues sooner.

Keep working while slides render

While pages generate, you can still edit brief, outline, or design. If one slide fails, it’s marked in the tray; the rest still complete, and you can fix that page afterward.

Great when you need to↓
  • A deadline is close and you want a full first draft fast
  • Long decks where waiting slide-by-slide isn’t an option
  • Review early pages while later ones are still rendering

Plans & background continuation

Concurrent batches per project depend on your plan: Free allows one; paid plans up to four. When batch-rendering from an outline, chat may finish first while rendering continues on the server—reopen the project to see thumbnail progress; the composer shows background rendering or queued status.