Ad Copy and Visuals in One Workflow: A Marketing Team's Guide.

Why copy and visuals keep breaking apart

Most marketing teams still write copy in one tool and generate visuals in another. Ad copy gets drafted in a chat window. Product shots or campaign imagery get generated somewhere else entirely. The two only meet properly in a deck, usually the night before a launch, when someone notices the headline promises "bold colour" and the hero image is desaturated grey. That's not a creative failure. It's a workflow failure.

Brand voice drifts fastest at exactly this seam. A copywriter working from a brief and a designer working from a moodboard are, in practice, drawing from different source material even when they sit two desks apart. The copy sounds like the brand. The visual looks like whatever the image generator defaulted to that day. Neither person did anything wrong, but the outputs don't feel like they came from the same campaign.

The real cost shows up mid-test. Performance ads live and die on iteration: a headline underperforms, you swap it, and now the paired visual needs to change too, because the hook and the image were never designed as a matched pair in the first place. That's the handoff where campaigns stall. Someone has to notice the mismatch, re-brief the visual, wait for a new export, and re-upload, all while the ad set is losing momentum in the platform's learning phase.

Marketing Studio (/marketing-studio) treats social posts and performance ads as one job, not two separate deliverables. There's no toggle between a "social" mode and an "ad" mode, and no separate export step between the line that sells and the image that shows it. Copy and visuals are generated in the same place, which means they can be iterated in the same place too.

Inside Marketing Studio: what it actually does

Marketing Studio generates carousels, ad formats and research-backed copy in the same studio, without forcing a switch between contexts depending on whether the asset is destined for an organic feed or a paid placement. That distinction matters less than it used to. A carousel built for a product launch might run organically this week and as a paid unit next week, with only minor copy adjustments. Building it once, in one place, means both versions come from the same source.

Copy generation draws on research-backed inputs rather than producing generic filler that reads like every other AI-written ad on the internet. Headlines and body text are shaped by category context and audience language, not just a prompt and a hope. Visual generation sits alongside that copy generation, so a campaign concept and its imagery develop together instead of copy arriving first and visuals being retrofitted around it.

A practical example: DTC skincare launch carousel

Take a product launch carousel for a direct-to-consumer skincare brand. The common approach, even in well-run teams, is to generate 10 to 15 image variations and 20 to 30 copy variations as separate pools, then let the ad platform's optimisation engine figure out which combinations perform. That's not creative direction, it's outsourced guesswork.

Working inside Marketing Studio, the headline variants and product shots get tested as paired sets from the start. "Skin that remembers what it needs" runs against a macro shot of the serum dropper mid-drip. "Three ingredients, zero guesswork" runs against a flat-lay of the product next to its three hero ingredients. Each pair is a deliberate creative decision, not a random combination assembled by an algorithm after the fact.

Research first, copy second: using Ray and Research together

Good ad copy starts before the first headline is typed. Research (/research), backed by Perplexity, pulls competitor positioning, audience language and category trends before anyone commits a line to the page. For a marketing team, that might mean surfacing how three competing brands frame the same product benefit, so the copy angle chosen is a genuine gap rather than an echo.

Ray (/ray) works as a creative director inside project chats. Ask it which writing model suits a punchy paid social hook versus a longer-form brand story, and it'll walk through the tradeoff rather than defaulting to one model for everything. A six-word hook built to stop a scroll has different requirements from a 400-word landing page narrative, and treating them the same is how brand voice gets flattened.

Drafting at scale with Write

For longer campaign briefs or landing page copy, Write (/write) gives you a six-model picker: GPT-5.6 Sol, Terra and Luna, Gemini Flash and Pro, and Claude Sonnet 5 and Opus 4.8, all on every plan. That range matters because the same brief drafted through different models surfaces genuinely different angles, not just paraphrased variations of the same idea. Run a launch brief through Claude Opus 4.8 for a considered, narrative-driven take, then the same brief through GPT-5.6 Luna for something faster and punchier, and compare.

Consider a game studio preparing a launch trailer campaign. The team drafts a dozen ad copy variants in Write, some leaning on atmosphere ("a world that remembers you left"), some leaning on mechanics ("build, break, rebuild, in real time"). The strongest line, once chosen, gets handed straight into Marketing Studio, where it becomes the anchor for the paired visual, rather than sitting in a doc waiting for someone to brief a designer separately.

From ad concept to published asset without leaving the workflow

Boards (/boards) is where visual direction gets collected and grouped before anyone commits to a final ad visual. Boards merges the old storyboard and moodboard functions into one fluid canvas, so a team can pull reference images, group frames into rough sequences, and settle on a visual direction before spending generation credits on a direction that won't survive review.

Once direction is set, actual asset production happens in Image (/generate/image) or Graphics (/graphics), depending on whether the format calls for photographic realism or vector-based graphic design. A product carousel for a skincare launch likely leans on Image for photorealistic bottle shots. A campaign built around bold typographic statements, the kind graphic designers reach for in poster-style social ads, leans on Graphics instead.

If the campaign needs a landing page to match, Web (/web) builds multi-page microsites with an SEO panel controlling title, description and social image per page, so the destination page tells the same story as the ad that drove someone there. Building the site works on any plan; publishing it live with a custom domain is a Pro and above feature, but the design work itself isn't gated behind a tier.

Cross-discipline example: exhibition design

An exhibition design team producing a trade show stand doesn't just need a stand render. It needs a promotional carousel to drive registrations and a microsite explaining the stand's concept to visitors booking meetings in advance. Traditionally that's two briefs to two different agencies, one handling social creative, one handling the web build, with brand assets shared by email attachment and hoped-for consistency. Producing both from the same brand assets in one workflow means the colour palette, typography and photography style actually match, because they came from the same source rather than two interpretations of the same brand guide.

Task	Traditional stack	In Marketing Studio
Ad copy	Separate chat tool	Research + Write, feeding Marketing Studio
Ad visuals	Separate image generator	Image / Graphics, same workspace
Landing page	Separate web builder, manual copy transfer	Web, sharing brand assets automatically
Testing	Copy and visuals as separate pools	Paired copy-and-visual variants

Keeping teams aligned: Projects and shared workspaces

Projects (/projects) is the shared workspace where brand identity, past campaigns and in-progress ad sets live together for the whole team. Instead of a copywriter working from one brand deck and a motion designer working from a different, slightly outdated version pulled from a Slack thread three months ago, everyone pulls from the same reference material, updated in one place.

For teams who want to build repeatable, traceable pipelines, Canvas (/canvas) is a node-based workflow editor where copy output can pipe directly into visual generation. A LlmChatNode drafts a headline, and that output feeds directly into an Image Generate node as part of the prompt, all in one traceable flow. A RayNode can sit upstream as an agentic planner, deciding how a campaign brief should branch into multiple copy-and-visual pairs before generation even starts. For teams producing multi-shot video ads, the Assemble Film node handles batch orchestration across several generated shots at once.

The practical payoff is fewer handoffs. Fewer handoffs means fewer versions of "final_v3_ACTUALFINAL_useThisOne" scattered across a shared drive, because there's no export-and-reimport step between the copy tool and the visual tool in the first place. The workspace that generated the headline is the same workspace that generated the image and, if needed, the landing page it links to.

Marketing teams don't need five subscriptions open in five tabs to keep copy and visuals in sync. They need one workspace where research, copy, imagery and the landing page it drives all draw from the same brand context, so a mid-test headline swap doesn't leave the visual behind.