Krea's Ad Studio Turns Any Brand Website Into a Full Ad Campaign

Krea has launched Ad Studio, a workflow that analyzes a brand’s website and generates advertising assets based on its visual identity. The product uses an ad-focused variant of Krea 3 and aims to remove much of the manual prompting typically required to produce a coordinated campaign.

Ad Studio is available through Krea’s app and is free to try. Krea says the app has more than 30 million users. The company has not announced separate Ad Studio pricing or a dedicated API. Its published Krea 3 inference rates are $0.03 for a 1K image and $0.06 for a 2K image, before retries, fine-tuning, or other workflow costs.

A URL becomes the brief

Ad Studio begins by scanning a website for logos, colors, typography, product imagery, and other visual cues. It uses those elements to assemble a brand kit, then generates variations intended to preserve the same products and identity across the set.

Users can correct or replace anything the website scan misses. Krea also provides a prepared Claude prompt that can construct the kit, giving teams another route when a site has incomplete assets or an unusual structure.

Krea 3 supplies the image-generation and editing layer beneath the workflow. Krea describes it as a foundation model for fast, targeted changes to existing images and reports end-to-end 2K editing times below eight seconds. That speed could support workflows that require dozens of sizes, compositions, or market-specific variants, although the performance figures remain vendor-reported.

Consistency drives the workflow

Advertising production depends on details surviving across every variation: packaging geometry, logo placement, brand colors, typography, and recognizable people or characters. General image models often drift on those details as prompts, layouts, and backgrounds change.

Krea 3 combines several controls designed to reduce that drift:

• LoRA fine-tuning: Teams can train a lightweight adapter on a particular product, person, or visual style. The adapter steers later generations without retraining the full model.
• Regional editing: Bounding boxes identify the area that should change, allowing the model to preserve the rest of the composition.
• Reference inputs: Each edit can use as many as six reference images, including product and style references.

For a product campaign, a LoRA adapter can learn the appearance of a specific SKU from approved photographs. Regional editing can then change the background, lighting, layout, or surrounding copy while reducing unwanted changes to the product itself. The output still requires review, especially where exact dimensions, labels, or regulated claims matter.

Where it fits in production

Krea has positioned Krea 3 for high-volume creative tasks that usually involve photography, compositing, and manual retouching. Its examples include turning snapshots into studio-style packshots, creating product colorways, localizing posters, extending a single concept into a series, and testing architectural materials or lighting.

Ad Studio packages those capabilities for marketing teams that do not want to assemble a workflow from Krea’s underlying image tools. Likely uses include:

• Expanding one hero image into multiple formats and aspect ratios
• Localizing a campaign for different languages and markets
• Producing product color variants without another photo shoot
• Creating batches of concepts for paid-social testing
• Maintaining a recurring mascot, spokesperson, or visual style

Developers evaluating the product should account for the current app-based access, the absence of announced Ad Studio API terms, and the review steps required before generated assets reach production. Listed inference prices also exclude the cost of rejected generations, asset preparation, fine-tuning data, and final design work.

Text remains a weak point

Independent testers have cautioned that Krea’s performance figures are vendor-measured and may exclude the time required to process reference images. Early outputs have also shown errors in small text, a recurring limitation across image-generation models.

Small-text failures create concrete production risks for prices, offer terms, legal disclaimers, product labels, and calls to action. Teams should plan to add final text in a conventional design tool and verify every factual, legal, and brand-sensitive element before publication.

Krea’s launch post also says Krea 3 has not reached real-time performance. Ad Studio therefore operates as a batch-generation workflow, with live 30-frames-per-second iteration outside the current release.

Krea packages the model around a job

Ad Studio combines model inference, brand ingestion, reference management, and structured outputs in one product. That packaging reduces the prompt engineering and orchestration work required from creative teams while giving Krea a workflow tied directly to paid media production.

A separate pricing analysis estimates that Krea 3 costs substantially less per image than some competing services. Per-image rates provide only a starting point for comparison; teams will need to measure usable-output rates, latency with references, fine-tuning quality, text accuracy, and human review time against their own campaign requirements.