The brief replaces the prompt in Pletor 2.0

With Pletor 2.0, a brief in natural language is all it takes to produce brand visuals and advertising content. Its creative agent relies on a Brain that centralizes identity, products, references, and campaigns.

“Create our back-to-school campaign around these two products, drawing inspiration from our best-performing content.” An instruction like this should now be enough to generate a series of on-brand visuals and ads, without manually selecting each model or rebuilding the production pipeline.

With version 2.0, Pletor describes itself as the first creative agent built for commerce. Users describe the expected result in plain language. The system interprets the brief, prepares the instructions, selects the necessary tools, and gathers the resulting assets on a canvas where the team can refine them.

The main change concerns how users enter the platform. Until now, production primarily relied on Flows, visual sequences connecting image, video, audio, or text models with brand rules and various operations. These systems remain available for projects requiring greater precision or intended for repeated production at scale.

The new agent handles the intermediate steps for more common requests. A team can ask for a product shot, an ad variation, or a complete campaign without choosing each component itself. According to Pletor, the same request can produce a single asset or a series of up to one hundred creations.

Personalization relies on Brain, the brand memory previously introduced by the company. It brings together identity, positioning, audiences, editorial rules, visual references, products, recurring characters, and commercial knowledge within a shared space.

According to the Brain documentation, this information is organized into short, hierarchical Memory entries. A company can provide its website address, brand guidelines, product sheets, previous campaigns, logos, images, documents, or audiovisual files. The system extracts relevant information, then asks the user to review and correct it.

Once built, this database can be accessed by conversations, Flows, Apps, and batch processing. A change made in Brain is reflected across productions using the same information. Agencies can also create a separate memory for each client.

The system functions more like a structured context database than a model trained specifically for each company. Content is still produced by external services, but those services receive the rules, references, and materials selected within Brain.

The platform’s catalog includes Nano Banana, GPT Image, Flux, Higgsfield, Reve, Seedance, Kling, Veo, Sora, Grok, and Claude. Users no longer necessarily have to determine which one is best suited to each stage. The agent is expected to distribute the work among the available tools based on the requested outcome.

The exact criteria behind this selection are not documented. The presentation does not explain how the system balances quality, speed, cost, and adherence to references. It also does not specify how many steps are performed for each brief or how a rejected creation influences the next attempt.

In this case, the term “agent” refers to an interface capable of turning a broad request into a creative production by coordinating several tools. The published information does not yet demonstrate full autonomy comparable to a system that could plan a campaign, measure its results, and independently revise its strategy. Teams retain responsibility for approving the content and can return to Flows when the automated process does not provide enough control.

The company also claims that Brain becomes more accurate with use. Each creation, decision, and successful campaign is said to strengthen its representation of the brand. The documentation explains how to add new references, approve extracted information, or edit a Memory entry, but does not clearly confirm that every choice made on the canvas automatically enriches Brain.

This distinction prevents an updatable memory from being confused with continuous learning. When a campaign performs better, adding it can guide future productions. Nothing yet establishes that the platform can independently access those results, understand why the campaign succeeded, or adjust its creative direction without human involvement.

Existing tools complement the new interface. Batch mode applies the same Flow to multiple products or variations. Apps turn a production sequence into a simplified form for a team or client. An API and MCP connection also provide access to Flows from external software or compatible assistants.

The business model relies on credits, with consumption varying according to the service used. Pletor’s pricing starts at $19 per month for 1,200 credits and one Brain. The Builder plan costs $49 for 3,300 credits and up to three Brains, while Studio costs $199 for 15,000 credits and up to ten Brains. Accounts can include unlimited users.

A free trial without a credit card provides a small number of credits, but Brain is listed as a paid-plan feature. Estimates expressed as numbers of images or videos remain approximate because consumption depends on the selected model, duration, quality, and format.

Storing a product catalog, internal campaigns, and brand documents also raises privacy concerns. In its data and intellectual property documentation, the company states that it does not use inputs, outputs, or Flows to train models without explicit consent. Users retain ownership of generated content while remaining responsible for securing the rights associated with uploaded materials.

Plans for larger organizations add advanced access controls, data hosting in the United States or European Union, and the ability to use private API keys. ISO 27001 and SOC 2 Type II certifications are mentioned as part of the Enterprise plan.

The central promise remains difficult to measure from the demonstrations alone. No independent test has yet compared Pletor 2.0 with a process based on conventional instructions. The company has not published first-pass approval rates, reductions in revision volume, or an evaluation focused on brand fidelity.

Adherence to a brand identity will necessarily depend on the quality of