Generate a campaign key visual from the CLI

Freeze the input contract before opening the CLI

Write the brief in fields that can be checked: campaign objective, generic subject, setting, camera distance, focal point, palette, lighting, material cues, intended copy zone, master ratio, and prohibited brand or legal elements. A useful example is: “16:9 campaign hero, sculptural amber glass form on pale stone, low three-quarter camera, warm side light, charcoal-to-cream background, subject on the right third, smooth empty gradient across the left third for later copy, realistic reflections, clean unmarked surfaces.”

Use the model ID returned by `npx @offlinecreator/mcp@0.1.1 models`; do not infer an ID from the display name. The version-pinned CLI requires `--model` and `--prompt`, recognizes `--aspect` or `--aspectRatio`, and forwards only `1:1`, `16:9`, or `9:16`. It has no dedicated CLI flag for a copy-safe rectangle, so the reserved area belongs in the positive visual description and remains subject to review.

Require a human gate before derivatives

Inspect the downloaded master at full size and at the intended placement size. Check subject integrity, unwanted text or marks, surface artifacts, edge continuity, lighting logic, protected-brand resemblance, and whether the reserved copy zone is genuinely low-detail. Place a temporary headline, subhead, and call to action over the proof in the design tool; reject the master if readable type requires covering the subject or if the crop depends on generative repair.

Record an explicit decision: approve, revise one named prompt variable, or stop. Approval covers this image and this campaign brief only. It does not prove the model will preserve the subject across another generation. Keep final copy, logos, accessibility text, disclaimers, and localization outside the generated pixels so those elements remain editable and reviewable.

Approve the 8-credit proof as a spend checkpoint

Run `npx @offlinecreator/mcp@0.1.1 balance` immediately before generation, then compare that balance with the live `models` response. The current public catalog lists FLUX Schnell at 1 credit for fast iteration, FLUX 1.1 Pro Ultra at 8 credits for photoreal detail, and Recraft V3 at 5 credits for brand graphics. Those are model-specific routing and cost facts, not quality rankings.

If composition is still unsettled, test broad directions with the lower-cost model before paying for detail. Once the art direction is fixed, authorize one Ultra master proof rather than three format variants. A failed job may return reserved credits under Studio's published policy, but a visually unsuitable completed image is still a completed proof; budget the review loop instead of assuming a refund for subjective rejection.

Proof model

FLUX 1.1 Pro UltraCurrent Studio fit: photoreal detail.

Published proof cost

8 creditsConfirm against the live models response and current balance.

Initial quantity

One 16:9 master proofApprove composition and copy space before any additional generation.

Separate upstream model capability from the CLI surface

Black Forest Labs documents FLUX 1.1 Pro Ultra as an upstream text-to-image model that can produce images up to 4 megapixels at varying aspect ratios. Its native API documentation also lists Raw mode and image-prompt controls. The OfflineCreator 0.1.1 CLI does not expose those native controls: its published implementation sends a model ID, prompt, and one of three recognized ratios. Do not add guessed `--raw`, seed, resolution, or reference-image flags.

OfflineCreator's model page says its launch examples are original AI-generated images produced through the same disclosed endpoints and displays a FLUX 1.1 Pro Ultra proof. That is real sample-existence evidence, but the retrieved page does not disclose the sample prompt, seed, or settings. It therefore cannot establish reproducibility, campaign suitability, or a success rate for this guide's prompt. The output from the authorized run is the only task-specific proof.

Supported conclusion

Higher-detail model fit and a published sample existStudio positions Ultra for photoreal detail; the provider documents up to 4MP upstream output.

Unsupported conclusion

The example is reproducible from this guideNo prompt, seed, or complete settings were available in the retrieved Studio page.

CLI boundary

Model, prompt, and three ratiosNative provider parameters are not automatically OfflineCreator CLI parameters.

Derive formats from the approved master, not fresh prompts

Create a crop map after approval. For 1:1, preserve the focal subject and one usable copy zone; for 9:16, define the vertical crop and safe top and bottom margins; for the original 16:9, retain the full composition. Add final typography as editable layers, then document each placement's crop, copy scale, and accessibility treatment. This deterministic derivative pass protects the approved direction better than asking the image model to reinterpret the campaign separately for every channel.

Return to the creative-workflow directory if the asset type is still undecided. Use the brand-graphic guide when the campaign needs a design-led graphic rather than a photographic hero. Use the social-variant guide when the next task is a controlled CLI generation experiment instead of cropping one approved master.

Keep the page inside its campaign-key-visual boundary

This page owns the narrow query “cli campaign key visual”: selecting a higher-detail image model, verifying its current credit cost, producing one reviewable master, and planning deterministic derivatives. It does not own broad workflow discovery, graphic-first brand artwork, or independent generated variants. Consolidate it into the parent workflow directory if current model fit, credit cost, and honest sample evidence can no longer support that narrow task.