AI Marketing Campaign Workflow: Brief to Full Creative Suite.

```html

Start With the Brief, Not the Blank Page

Most campaign delays don't happen in production. They happen in the gap between a vague brief and the moment someone finally commits to a direction. Closing that gap quickly is where an AI-assisted workflow earns its keep.

The first move is grounding the brief in real data rather than assumptions. Stensyl's Research surface (/research), backed by Perplexity, lets you ingest competitor positioning, platform trends, audience language, and category signals before writing a single word of copy. A graphic designer briefing a rebrand campaign, a product team planning a launch, a content studio preparing a social series: all of them benefit from starting with what the audience actually cares about, not what the client thinks they care about.

Once you have your research grounding, open a Project (/projects). This is where brand identity files, tone-of-voice documents, reference decks, and team notes live. Every surface downstream, from Write to Video to Marketing Studio, pulls from the same project context. Nothing gets re-explained. Nothing drifts.

Before touching any generation surface, spend five minutes with Ray (/ray). Describe the campaign objective: the audience, the channel mix, the business goal. Ray will recommend which surfaces and models to use in sequence. That recommendation saves the planning time most teams burn on tooling decisions rather than creative ones. Ray runs on Claude Sonnet or Opus with web search, so its recommendations are grounded, not generic.

For campaigns that will run more than once, or need multiple variants across channels, take one extra step. Set up a Canvas (/canvas) workflow with an LlmChatNode feeding campaign parameters into downstream generation nodes. The brief becomes a reusable pipeline rather than a one-off document. Change the audience segment, update the product, and the whole chain re-runs. That is what separates a campaign workflow from a collection of tasks.

Writing Campaign Copy Across Formats

Campaign copy rarely needs to be a single document. It needs to be a landing page, six email subject lines, three paid social headlines, a DM script, and a product description, all saying the same thing in different registers. That is the practical case for a multi-model copy workflow.

Stensyl's Write surface (/write) gives you a multi-model picker inside a single long-form drafting environment. The six models available on every plan serve distinct purposes within a campaign stack:

• Claude Opus 4.8 is the right choice for persuasive narrative, brand voice, and long-form formats: landing pages, nurture email sequences, video scripts, and brand manifestos. It reasons through tone carefully and follows complex brand instructions reliably.
• Gemini Flash is built for speed and volume. Use it to generate twenty subject line variants, thirty social captions, or rapid keyword-aligned snippets for performance channels. It won't match Opus on depth, but it produces usable variants at a pace no human writer can match.
• GPT-5.5 handles structured ad copy with precision: headlines with character counts, CTAs formatted for platform specs, schema-driven copy blocks for paid search and display. Its output tends to be tighter and more format-obedient than the Anthropic models, which matters when Google Ads or Meta has strict field limits.
• Claude Sonnet 4.6 sits between Opus and Flash: strong enough for mid-length campaign copy, fast enough for iterative drafting.
• Gemini Pro is useful for copy that needs to reflect search intent closely, particularly for SEO-led landing pages or content written to match keyword clusters from the Research phase.
• GPT-5.4 mini is the right model for bulk low-stakes copy: product description variants, meta descriptions, and alt text at scale.

The practical workflow: draft the hero copy in Opus, then switch models to generate variants without leaving the Write session. Compare a Gemini Flash subject line against an Opus version in the same document. Choose. Move on.

For campaigns running across multiple channels simultaneously, pipe copy briefs through Canvas LLM Chat nodes to batch-produce variants in parallel. One node drafts the email sequence, another handles the paid social copy, a third generates the organic captions. All draw from the same campaign parameters defined at the brief stage.

Before finalising body copy that makes factual claims, including statistics, product comparisons, or market-share figures, run a Research pass to verify. This keeps the campaign legally defensible and accurate, which matters particularly for regulated sectors like financial services, health, or automotive.

Building the Visual Layer: Static and Graphic Assets

Campaign copy without a consistent visual language is half a campaign. The visual layer has to be built in the same project context as the copy, not assembled separately and merged at the end.

Start in Boards (/boards). Before committing to a full asset batch, pin reference images, annotate them for the team, and define the visual direction. Boards merges moodboarding and storyboarding into a single fluid canvas, so a motion designer briefing an animated banner campaign and a product designer defining a hero image aesthetic can both work in the same space. Once the visual language is agreed, generation becomes execution rather than exploration.

Move to the Image surface (/generate/image) for hero asset generation across 20-plus models. The model choice matters:

• Luma Uni-1 handles typography inside images well, making it the right choice for campaigns where a headline or product name needs to read cleanly as part of the visual rather than being added in post. Exhibition designers producing event graphics and graphic designers building typographic campaign posters will find this particularly useful.
• For product-led campaigns, particularly in automotive, consumer goods, or packaging, use Scene Composer (/studio/compose). Pose 3D product models against branded backdrops, adjust lighting and camera angle, and render photorealistic hero shots without booking a studio. The output is campaign-ready imagery at a fraction of the cost and lead time of a physical shoot.

Once you have hero images, move to Graphics (/graphics) for vector-based deliverables. Social carousels, display banners, and print-ready layouts need precise compositional control that image generation alone doesn't provide. Graphics gives you that environment, letting you take the visual direction established in Boards and translate it into production assets with correct dimensions, bleed, and layering.

The final step before export is Editing (/editing). Use it for background removal, upscaling for large-format print, and final retouching. For a retail campaign delivering assets across digital display, out-of-home, and social, the same hero image often needs to exist at five different specifications. Editing handles that without re-generating from scratch.

Asset Type	Surface	Model / Tool
Campaign hero image	Image (/generate/image)	Luma Uni-1 for text-in-image
Product packshot / 3D render	Scene Composer (/studio/compose)	3D model + branded backdrop
Social carousels / display banners	Graphics (/graphics)	Vector + layout generation
Visual references / scene planning	Boards (/boards)	Fluid canvas, annotated references
Background removal / upscale / retouch	Editing (/editing)	Inpaint, upscale, BG removal

Adding Video: From Single Clips to Full Campaign Reels

Short-form video is now table stakes for most campaign channels. A product launch needs a teaser clip. A brand campaign needs a 30-second social cut. A direct-response ad needs a talking-head presenter. Building all three in the same platform, without re-briefing across separate tools, is where a consolidated video workflow saves real time.

Start in Boards. Define the start and end keyframes for each scene before generating any video. This gives you first-frame and last-frame control from the outset, which is critical for brand consistency across a multi-clip campaign. Boards handles this natively, so the scene plan feeds directly into video generation.

For individual clips, use the Video surface (/generate/video). The two primary Luma models serve different purposes:

• Luma Ray 3.2 is the right choice when frame accuracy matters: a product shot that needs to land on a specific end-frame call to action, or a sequence where the first frame must match the last shot of the previous clip. It supports 5- or 10-second outputs with start and end keyframe control, and looping for social formats.
• Luma Ray 2 Flash is for rapid low-cost draft iterations. Test motion concepts, check whether a composition reads well in motion, and validate camera moves before committing credits to a final Ray 3.2 render.

Once individual clips are approved, move to Film (/film) to assemble multi-scene campaign videos. Film lets you sequence clips into a coherent narrative with consistent framing and pacing. A game studio building a launch trailer, a fashion brand assembling a seasonal campaign reel, or an automotive team producing a model reveal: all of them need the same thing from Film: a way to turn approved individual clips into a sequenced whole without re-exporting and manually editing in a separate timeline tool.

For direct-response formats, particularly paid social ads where a presenter-to-camera format outperforms pure motion content, use Avatar (/avatar). Train a digital twin on approved footage, apply branded outfits, and render talking-head videos for performance campaigns without booking a shoot. This is particularly useful for localisation: one script in five languages, rendered by the same digital presenter, produced in hours rather than the days a multi-location live shoot would require.

After assembly, use Editing (/editing) to add captions. Whisper speech-to-text transcribes the audio accurately, karaoke mode lets you style and position the captions, and they bake directly into the final MP4. For social platforms that autoplay without sound, baked captions aren't optional: they are the difference between a video that performs and one that doesn't.

Assembling Ads and Social Deliverables in Marketing Studio

Most campaign teams produce final ad creative in a different tool from the one they used to write the copy and a third tool from the one they used to generate the images. The asset transfer between those tools is where briefing gets lost, formats get wrong-sized, and approvals stall.

Marketing Studio (/marketing-studio) is a single surface for social posts and performance ads. Carousels, ad formats, and research-backed copy are all built here, without toggling between separate social or ads environments. There is no separate Ads Studio or Social Studio: it is one surface, and it handles both.

The practical advantage is in the connections it inherits. Pull copy directly from Write and visuals directly from Graphics into Marketing Studio to assemble finished ad creative without re-exporting or reformatting assets. The copy was written with the correct model for the format. The visual was produced to the right dimensions. Assembly in Marketing Studio is assembly, not re-production.

For ads that need movement, use Motion (/motion) to add animated elements to static ad frames. Motion uses Remotion under the hood and exports production-ready motion graphics files suitable for paid placements. A graphic designer building an animated display banner for a brand campaign, or a motion designer producing a looping social ad for a product launch, both work here rather than exporting to a separate animation tool.

Video ads follow the same pipeline: render in Film or Video, caption in Editing, then bring the finished MP4 into Marketing Studio as the creative asset. The caption step is worth repeating: baked captions in the MP4 mean the file is platform-ready on export, with no dependency on platform-side auto-captioning which varies significantly in accuracy across networks.

Running the Full Workflow Without Switching Platforms

The value of a connected workflow isn't just convenience. It is that context doesn't get lost between stages. When the brief, the copy, the visuals, and the video all live in the same project, downstream decisions stay grounded in upstream decisions. The copy reflects the research. The visuals reflect the copy direction. The video reflects the visual language. That coherence is what makes a campaign feel like a campaign rather than a collection of assets that happen to share a colour palette.

Canvas (/canvas) is the connective tissue. Chain Research outputs into Write, Write into Image generation, and Image into Film using Assemble Film nodes for batch multi-shot orchestration. An LlmChatNode feeds campaign parameters downstream, so each generation stage inherits the context of every stage before it. What would otherwise be five separate briefs handed off between five separate tools becomes one node graph that reruns when the brief changes.

The credit model reflects this. On a Pro plan (6,000 credits, £42 per month), a full campaign cycle, research, copy variants, image generation, video clips, and ad assembly, runs on a single plan with a single credit pool. Compare that to the typical alternative: a copywriter LLM subscription, a separate image generator subscription, a video generation tool, an avatar platform, and a project management layer. Five subscriptions, five billing cycles, five sets of context that don't talk to each other.

The credit difference is real, but the context loss is the more expensive problem. When a social media manager briefs a new image in a standalone image tool, they re-explain the campaign from scratch. When they do it inside a Stensyl Project where the Research output, the copy brief, and the brand files already live, they don't. The brief is already there.

Projects (/projects) enforce this. Every asset version, brand file, and team comment lives in one shared workspace. Handoffs between copywriters, art directors, and motion designers happen inside the same environment. Version control is implicit rather than managed through email chains and shared drive folders with names like "FINAL_v3_USE_THIS_ONE".

For teams starting out, the Free tier includes 150 one-time credits and one free video render, with every model available. It is enough to run through the full pipeline once and verify that the workflow fits before committing to a paid plan. Credits on the Free tier don't reset, but they are sufficient to test Research, Write, Image, Video, and Marketing Studio in a real campaign context.

"The brief, the copy, the visuals, and the video all live in the same project context, so nothing gets lost in translation between tools."

The practical upshot is this: a campaign that used to require coordinating five tools, five logins, and five sets of exports now runs in a single environment from the first research query to the final baked-caption MP4. The creative decisions are still yours. The context just travels with them.

```