Canva AI Prompts: 36 Templates for the Tools That Take One

TL;DR: Roughly half the tools people call "Canva AI prompts" have no prompt box at all. This page sorts Canva's current AI surfaces into the ones that read your text and the ones that only read a click, then gives 36 copy-paste templates for the first group, organised by the job.

Does Magic Studio still exist in Canva?

Not as a place you can go. Checked on 29 August 2026, canva.com/magic-studio/ returns a 404, and the shortlink canva.com/magic redirects to https://www.canva.com/canva-ai/. Canva's own site navigation now carries a section headed Canva AI, listing Magic Layers, Canva AI assistant, Canva Code, Magic Resize, Magic Animate, Magic Write, Translate, Magic Insights and Magic Formulas. The phrase "Magic Studio" appears on none of it.

The name survives in exactly two places. One is a legacy help article, /help/using-magic-studio-safely-and-legally/, whose own heading reads "Using Magic Studio Responsibly" and which still describes Magic Expand and Magic Edit in the old vocabulary. The other is the tool names themselves: Magic Media, Magic Write, Magic Edit, Magic Eraser and Magic Expand are all live features with current help pages.

This matters practically. Most "Canva AI prompts" content on the web describes menus that have moved, and a guide telling you to open Magic Studio from the homepage is describing a build that no longer ships. The umbrella is Canva AI, and since Canva Create 2026 the flagship is Canva AI 2.0, which Canva's help centre describes as a conversational partner rather than a one-shot generator.

Which Canva AI features actually take a prompt?

The question every listicle skips, and the one that decides whether a template is useful. From Canva's own help articles, read 29 August 2026:

Half the table is click-to-apply. No prompt engineering improves a Magic Expand result, because Magic Expand never sees words. A "Canva prompt pack" offering templates for erasing objects or expanding a photo was written by someone who has not opened the tool.

What does a Canva prompt have to carry that a chat prompt does not?

The design type, the audience, the outcome, and the visual register. Canva's own prompting guide is unusually specific about this. Its AI-powered design section tells you to start with the design type you want, be clear on the end goal and who it is for, specify the style, and outline the key message. Its image section says to begin by specifying the image's purpose. Its video section says to put the main subject or action first, and adds a syntax note most people miss: "Use brackets or parentheses to call attention to important details you don’t want missed."

That is a different shape from a chat prompt. A chat prompt asks for text you will then edit; a Canva prompt asks for a finished artefact on a fixed canvas, and every decision you leave open gets filled by a template default. A workable Canva brief carries six slots:

1. Artefact — Instagram post, 16:9 presentation, A5 flyer, 8-second clip.
2. Audience — who reads it, in five words or fewer.
3. Job — what it has to make them do.
4. Message — the one line that must survive.
5. Register — colour, mood and style words, not adjectives about quality.
6. Constraint — brand kit, language, length, what to leave out.

Every template below is that skeleton with the variables exposed. Replace anything in [SQUARE BRACKETS] and paste.

Prompts for social posts and captions

Canva AI generates social designs directly. Its help page lists Instagram posts and stories, Facebook posts and covers, Pinterest pins, YouTube thumbnails and banners, and X/Twitter posts among the supported types.

1. SINGLE SOCIAL POST
Create an [INSTAGRAM POST] for [AUDIENCE] promoting [OFFER].
The one thing it must communicate is [MESSAGE].
Style: [WARM AND FRIENDLY / CLEAN AND MODERN / BOLD], using [COLOURS].
Leave room for a logo in the [BOTTOM RIGHT].

2. CAPTION FROM A FINISHED DESIGN  (Ask Canva, page selected)
Write a caption for this design aimed at [AUDIENCE].
Tone: [TONE]. Length: [UNDER 40 WORDS].
Open with the benefit, not the product name. End with [CALL TO ACTION].
Do not use hashtags.

3. MULTI-PAGE CAROUSEL
Create a [6]-page [INSTAGRAM CAROUSEL] explaining [TOPIC] to [AUDIENCE].
Page 1 is the hook, pages 2 to [5] are one idea each, page [6] is the ask.
Keep every page to [12] words or fewer of on-image text.
Style: [STYLE]. Colours: [COLOURS].

4. PLATFORM VARIANTS OF ONE MESSAGE
Take this message: "[MESSAGE]".
Write [3] caption variants for [LINKEDIN], [INSTAGRAM] and [X].
Match each platform's normal length and register.
Keep the claim identical across all three; change only the framing.

5. POST BUILT AROUND MY OWN PHOTO
Create an [INSTAGRAM STORY] using the attached photo as the main image.
Audience: [AUDIENCE]. Message: [MESSAGE].
Add a headline of [6] words or fewer and a [CALL TO ACTION] at the bottom.
Do not crop faces. Keep text clear of the top and bottom [15%].

How do you get a usable presentation from one brief?

By making the outline decisions before Canva invents them. Canva AI generates a presentation from a prompt, then shows an editable outline where you can add, delete and reorder sections before slides exist. That step only appears when you generate from the homepage; Canva's help page notes the Templates tab in the editor shows fewer options.

A second route: Edit with AI sits on template cards, Brand Template cards and your own designs in Projects, and generates a design following that template's style plus your prompt. Canva documents it for presentations, social media and print designs only.

6. DECK FROM A BRIEF
Create a [12]-slide presentation on [TOPIC] for [AUDIENCE].
Argument: [THE ONE THING THEY SHOULD BELIEVE AT THE END].
Slide 1 is the problem, slides 2 to [10] build the case, slide [11] is the ask.
Keep each slide to one idea and [25] words or fewer.
Style: [PROFESSIONAL AND MODERN], colours [COLOURS].

7. DECK FROM AN ATTACHED DOCUMENT
Using the attached [BRIEF / REPORT], build a [10]-slide deck for [AUDIENCE].
Preserve the figures exactly as written; do not recalculate or round anything.
Turn each section heading into one slide.
Add a summary slide at the end listing the [3] decisions we need.

8. DECK FROM DATA
Create a presentation from this data for [AUDIENCE].
Chart [METRIC] over [PERIOD] as a [LINE CHART], one chart per slide.
Label every axis with its unit. Write a one-sentence takeaway under each chart.
Do not add slides that are not supported by the data.

9. EDIT WITH AI, FROM A TEMPLATE OR BRAND TEMPLATE
Keep this template's layout, fonts and colour system exactly.
Replace the content with [TOPIC] for [AUDIENCE].
Headline: [HEADLINE]. Body: [KEY POINTS].
Do not add new pages or change the image placement.

10. REFINE AN EXISTING DECK  (Canva AI panel, in the editor)
Review this deck as [AUDIENCE] would.
Then, one change at a time: [MAKE ALL HEADLINES ONE SIZE LARGER].
After that, [CONVERT EVERY PHOTO TO BLACK AND WHITE].
Stop after each change and tell me what you did.

11. SHEETS: BUILD THE DATA FIRST
Create a [BUDGET TRACKER] with columns for [DATE, CATEGORY, AMOUNT, NOTES].
Format [COLUMN C] as [US DOLLAR CURRENCY WITH 2 DECIMAL PLACES].
Sort [COLUMN A] alphabetically.

12. SHEETS: SUMMARISE FOR THE DECK
Summarise this sheet and highlight the main trends.
Then create a chart showing [MONTHLY SPENDING TRENDS].
Give me [3] sentences I can paste onto a slide.

Which image surface should you prompt?

Three answers, not interchangeable. Magic Media is the in-editor app: Images, Graphics and Videos tabs, style presets (photography, digital art, fine art; simple, artistic, detailed) and a square, landscape or portrait layout. Dream Lab is the standalone generator: four options per prompt, optional reference image, follow-up edit instructions, quality tiers and a resolution boost. Magic Edit is inpainting: brush or click a region in a photo you already have, then describe the replacement.

Canva's image prompting guidance is short and worth following literally: name the purpose, then describe the style, colours or mood. Its own examples read like art direction, not object lists.

13. MAGIC MEDIA IMAGE
[A BRIGHT SKINCARE SCENE] with [FRESH, DEWY TEXTURES] and [MINIMAL PROPS],
[CALM SETTING], [SOFT DAYLIGHT FROM THE LEFT], [PALE GREEN AND CREAM].
Style: Photography. Layout: [SQUARE].

14. MAGIC MEDIA GRAPHIC
A [FLAT VECTOR ICON] of [SUBJECT], [SINGLE COLOUR ON TRANSPARENT],
[THICK EVEN STROKES], [NO TEXT], [NO GRADIENT].
Style: Simple.

15. DREAM LAB, DETAILED SCENE
[SUBJECT] in [SETTING], [TIME OF DAY], [LIGHTING],
shot [WIDE / CLOSE], [MOOD], [COLOUR PALETTE], [NO PEOPLE].
Purpose: [HERO IMAGE FOR A LANDING PAGE].

16. DREAM LAB, FOLLOW-UP EDIT
Make the [CAR] [BLUE].
Keep the composition, lighting and background exactly as they are.

17. MAGIC EDIT, INPAINT
[Brush the region first, then:]
Replace this with [OBJECT], [MATERIAL / COLOUR],
lit from the [SAME DIRECTION AS THE REST OF THE PHOTO],
[SAME SCALE AS THE OBJECT BESIDE IT].

18. BG GENERATOR
[OPEN SPACE WITH BRIGHT STARS AND A DISTANT PLANET],
[HARSH LIGHTING FROM THE UPPER LEFT], [COOL BLUE TONES].
Describe the scene, not the subject. Keep the subject's edge lighting consistent.

Can you prompt text effects and headlines?

Only the headlines. Canva's text effects are a closed list of presets applied from a panel — Drop, Glow, Echo, Outline, Background, Splice, Hollow, Glitch, Neon, Curve — with sliders for intensity, colour, direction and transparency. There is no text field in that flow, and TypeCraft, the app usually recommended as the "AI text effect" route, is the same: pick a font, style and colour, then drag control points. So the prompt-shaped part of typography here is the copy, plus the decorative elements around it.

19. HEADLINE SET  (Magic Write)
Write [8] headlines for [ARTEFACT] aimed at [AUDIENCE].
Each must be [6] words or fewer and contain [KEY TERM].
Half should lead with the benefit, half with the tension.
No colons, no wordplay, no exclamation marks.

20. HEADLINE THAT FITS THE BOX  (Magic Write, "This but…")
This but [4] words shorter, same meaning, same first word.

21. GENERATE A HEADLINE ELEMENT  (Canva AI 2.0, in the editor)
Add a headline to this page that says [MESSAGE].
Match the existing font stack and the darkest colour already on the page.
Size it so it fills the top [third] without touching the [PHOTO].

22. DECORATE, TO STYLE AROUND THE TYPE  (Edit Page panel)
[MINIMAL GEOMETRIC SHAPES] in [TWO BRAND COLOURS],
kept to the [OUTER MARGINS], nothing over the headline.

How do you keep generated copy in your brand voice?

Two mechanisms, and they work at different levels. Magic Write has a saved tone of voice: paste sample text, click Analyze Voice, name it, and Canva stores it. Its help page limits you to three saved voices. Canva AI 2.0 has Memory Library, which builds an "About Me" page from your designs and conversations and applies it to future output.

Both are approximations. Canva's writing-assistant troubleshooting page suggests that if the tone feels off, you check whether Brand Voice is affecting the output and turn it off temporarily to compare. Treat a saved voice as a nudge, and keep naming the tone in the prompt anyway.

23. SUPPLY THE VOICE SAMPLE  (Magic Write > Change voice > Add a voice)
[Paste 200-400 words of your own published copy. Pick text that is
typical rather than best: an average email beats an award-winning one.]

24. BRAND-VOICE COPY BLOCK
Write [BODY COPY] for [ARTEFACT], [LENGTH].
Voice: [SAVED VOICE NAME].
Rules: [SENTENCES UNDER 18 WORDS], [NO EXCLAMATION MARKS],
[BRITISH SPELLING], [NEVER SAY "SOLUTIONS"].
Audience: [AUDIENCE]. They already know [ASSUMED KNOWLEDGE].

25. TONE-CONSTRAINED CAPTION SET
Write [5] captions for [CAMPAIGN] in [SAVED VOICE NAME].
Each under [30] words. Each must open with a different sentence shape.
Reject anything that sounds like a press release.

26. VOICE AUDIT ON EXISTING COPY
Read this copy. List every phrase that does not sound like [SAVED VOICE NAME].
For each, give one replacement in the right voice.
Do not rewrite the whole thing; give me the diff.

A voice saved inside Canva exists only inside Canva. If the same voice has to hold across Canva, a chat model and your email tool, it needs to live somewhere portable — the point of our post on nailing your brand voice and the brand-voice context approach.

How do you repurpose one design across formats?

Not with Magic Resize, which is a size picker. The prompt route is Canva AI 2.0 in the editor, which Canva's help page lists as able to resize or flip a design, apply bulk edits across all pages, and create style variations on new pages.

27. RESIZE AND REFLOW BY CONVERSATION
Resize this design to [1080x1920].
Reflow the text so nothing is cropped and nothing overlaps.
Keep the headline the largest element. Move the [LOGO] to the [BOTTOM CENTRE].
Tell me what you moved.

28. BULK EDIT ACROSS PAGES
Across every page: [CONVERT ALL IMAGES TO BLACK AND WHITE].
Do not change any text, layout or colour other than the images.

29. FORMAT SWAP FOR ONE PLATFORM
Adapt this [PRESENTATION SLIDE] into a [LINKEDIN POST GRAPHIC].
Keep the claim and the chart. Drop the footer and the slide number.
Increase the type size so it is readable at thumbnail scale.

30. STYLE VARIATIONS TO CHOOSE FROM
Create [3] style variations of this page, each on a new page.
Vary only the colour system and the type pairing.
Keep the layout, the copy and the image positions identical.

Canva's own guidance is worth repeating: start with simpler designs, and break large requests into smaller steps. For turning one asset into a dozen, our content repurposing workflow covers the sequencing.

What do Canva video prompts look like?

Short, and split across two tools. Canva AI's Video clip generates 8-second videos with audio or 6-second videos without, and its help page states plainly that "Prompts are supported in English only". Magic Media's Videos tab is separate, and Canva documents that "Every generated video will be 4 seconds long." Neither is a script-to-film pipeline. You are prompting a shot, then assembling shots in the editor.

Accessing Video clip through Canva AI on the homepage also exposes Style, Aspect ratio, Framing and Lighting preferences, which is the closest thing Canva has to camera controls.

31. VIDEO CLIP, 8s WITH AUDIO
[SUBJECT AND ACTION FIRST], [SETTING], [TIME OF DAY].
Camera: [SLOW PUSH IN].
((Keep [DETAIL] sharp and centred.))
Mood: [MOOD]. Lighting: [LIGHTING].

32. VIDEO CLIP WITH A CAMERA MOVE
Begin with a [WIDE SHOT] of [SUBJECT] in [SETTING],
move closer to [THE SUBJECT'S FACE], then [PAN RIGHT] to show the surroundings.
[GOLDEN HOUR], [HANDHELD].

33. B-ROLL
Create b-roll footage of [SUBJECT], [MOVEMENT], [ENVIRONMENT].
No people, no text, no logos.
[LOOPABLE], [SHALLOW DEPTH OF FIELD].

34. MAGIC MEDIA VIDEO (4s)
[ONE ACTION ONLY], [SUBJECT], [SETTING], [LIGHTING].
Keep the frame static.

35. MUSIC BED
[CINEMATIC], [TENSE], [SLOW BUILD], [PIANO AND STRINGS].
Use: [LOOPABLE BACKGROUND UNDER A VOICEOVER].

36. SOUND EFFECT
[SOFT WHOOSH TRANSITION], [SHORT], [SUBTLE], [INDOORS].

Canva documents real content rules on the audio generator: no copying or recreating copyrighted songs, no audio in the style of a specific artist where the intent is to imitate, no protected brand music. If a request is blocked, Canva's advice is to describe mood, genre, instruments and tempo instead. Our breakdown of the seven parts of a video prompt maps that vocabulary onto models that accept more of it.

How much can you actually generate on your plan?

Canva publishes this, which is more than most vendors do. Its AI usage page splits tools into AI design tools that sit outside the allowance under fair use, then Standard, Premium and Ultra tools that all draw on one shared monthly pool. As the prompting guide puts it: "All premium AI tools share a single monthly allowance."

The published monthly ranges, read 29 August 2026:

Read those as ceilings, not budgets. Canva's own caveat is that each range assumes one tier only, a simple prompt, and the least demanding tool in that category. Canva AI 2.0 sits in the Ultra tier and consumes more when a task references large designs, takes more steps, or calls several tools. Allowance resets on your billing date for paid plans, and at 12:00 a.m. UTC on the first of the month on Canva Free.

Two more documented facts change how you plan. The AI Product Terms state that "AI Usage Limits are operational controls and not fixed entitlements", so the numbers above can move. And the quality-tier page notes that Ultra is paid-plans-only, that videos never run at Standard because the tier switches to Premium automatically, and that graphics, audio, 3D, forms, code, charts and sheets are Premium-only whatever you select. We are not quoting Canva prices, which vary by region and promotion; check Canva's pricing page.

Can you use Canva AI output commercially?

Canva's answer is yes with a carve-out, and the carve-out is the part people miss. The AI Product Terms, effective 26 June 2026, say that "you retain your ownership rights to your Input, and you own your Output, except for any Output that modifies or incorporates Licensed Content", and that you "may use your Output for any lawful purpose". The worked example is unambiguous: "If you use an AI Product to edit a photo from the Canva library, you won’t own this Output;". Generate a whole presentation containing library photos and you own the presentation, not the photos, which stay under Canva's content licence.

Ownership is also not exclusivity. Canva states that "outputs may not be unique, and other users may receive similar Outputs from our AI Products", and its responsible-use page goes further: "Currently, many jurisdictions (including the US) do not give copyright protection to AI-created works." That page warns against using AI output as a trademark for the same reason, and instructs you: "Don’t reference known brands, characters, or people in your prompts (unless you have express permission)."

Where Canva AI stops

Canva's AI is a design-tool assistant, not a general model, and that cuts both ways. Running inside templates, brand kits and a fixed canvas makes it unusually good at producing something on-brand, correctly sized and immediately editable. A general image model gives you a picture; Canva gives you a layout with movable layers. For a team that needs forty on-brand assets this week, that constraint is the value.

The same constraint is the ceiling. Output tends toward the house style, because the house style is what the system is optimised to reproduce, and Canva says so itself: it cannot guarantee your AI-created work will be unique to you. Magic Write's help page adds two limits no prompt clears, a 1500-word input cap and around a 2000-word output cap, plus a note that "It only uses data and information up to mid-2021" — Canva's statement about that tool specifically, not about Canva AI 2.0, whose web research feature exists to work around it.

The honest positioning for us: Canva is not on our integrations list, and Prompt Architects does not generate designs, images or video. What it does is write and store the prompt, which is the part that moves between Canva, ChatGPT, Claude and everything else you use. Our free plan includes 5 prompt enhancements per day, forever, per our FAQ page. If a template above is one you will run fifty times with different variables, saving it with variables beats keeping a document of them.

Sources and access dates

Everything above was read from Canva's own properties on 29 August 2026, over HTTP/1.1 with a desktop browser user agent, because canva.com/help/* returns a 403 over HTTP/2.

• Prompting and allowances: /help/ai-prompting-tips/, /help/ai-access/, /help/canva-ai-quality-tiers-standard-premium-ultra/
• Canva AI 2.0: /help/using-canva-ai/, /help/new-canva-ai/, /help/canva-design-assistant/
• Generators: /help/using-magic-media/, /help/generate-with-dreamlab/, /help/using-magic-edit/, /help/bg-generator/, /help/video-clip/, /help/generate-ai-music-sound-effects-and-voiceovers/
• Click-only: /help/using-magic-expand/, /help/magic-eraser/, /help/text-effects/, canva.com/magic-layers/
• Rights: canva.com/policies/ai-product-terms/ (effective 26 June 2026), /help/using-magic-studio-safely-and-legally/

Two inconsistencies inside Canva's own docs are worth knowing. The conversational-design article lists Pro, Business and Enterprise; the "What’s new in Canva AI 2.0" article lists Pro, Teams, Business and Nonprofit. Enterprise is in one only, Teams and Nonprofit in the other only. And the Docs writing assistant lists Canva Free as supported while the general 2.0 pages cap Free at 1.0 features. Check the article for the exact surface you are opening.