First AI visuals

How can you obtain an image suited to a simple need and identify its limitations before using it?

1250 € per session, excl. VAT

In short

This half-day session of 3 hours explains how AI produces images from instructions. Participants experiment with formulating a request, compare several results and select a visual for a fictional case. They leave with a documented image, a quality assessment grid and points to consider before distribution.

Why this training?

A plausible image does not necessarily match the request: distorted objects, illegible text or inconsistent details can go unnoticed. Understanding these limitations allows you to look at the result from more than just an aesthetic perspective. Generation also raises questions about rights, consent and misleading representation. Distinguishing between experimentation and publication from the very first attempts avoids treating a generated image as automatically usable. Ongoing case study: The fictional association Avelune Créations is preparing a poster for a pottery workshop for adult beginners. It is looking for an illustration showing clay objects on a table, with no identifiable person and a clear area for adding the title later. Application at work: For a future visual request, reuse the brief and the grid, use only fictional or anonymised data, then have the rights and notices checked and approved before any actual distribution.

Objectives of this training

  • Explain the difference between image generation, image searches and retouching in simple terms.
  • Write a prompt specifying the subject, composition, atmosphere and constraints.
  • Compare two results using a shared assessment grid.
  • Identify visual flaws and departures from the brief.
  • Identify the checks on rights, consent and notices required before distribution.

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Staff wishing to explore visual creation without prior AI experience.
  • Staff with occasional communication responsibilities who need to select simple illustrations.

Prior skills

  • No prior AI knowledge required.
  • Ability to use a browser and save an image file.

What you will work on

  1. What AI actually produces

    Objective: Distinguish between plausible creation and reliable representation. Learning from examples, text-guided generation, variability of results; differences from an image library and a photograph.

  2. From brief to first prompt

    Objective: Translate a visual intention into an explicit request. Main subject, framing, palette, lighting, descriptive style and space reserved for text; constraints that are sometimes imperfectly followed.

  3. Compare and correct without guesswork

    Objective: Measure quality, then adjust a request. Four-dimensional assessment grid; examination of textures, perspectives and objects; changing only one intention at a time, without expecting an identical reproduction.

  4. Preparing for responsible use

    Objective: Identify the conditions for using the selected visual. Terms of use, third-party rights and imported references; consent of the people depicted; notices according to the context and applicable rules; deepfakes and the risk of deception.

What you take away

  • Visual brief for the fictional poster.
  • Log of prompts and adjustments.
  • Visual selected for the mock-up, accompanied by its limitations.
  • Completed quality assessment grid and pre-distribution checklist.

Method and assessment

  • Short teaching inputs followed by immediate practice on the ongoing case study.
  • Individual work followed by work in pairs, with peer review of the outputs.
  • Guided group review: what works, what needs correcting and what remains a human decision.
  • Preparation: Check access to a tool authorised by your organisation and the applicable terms of use. Prepare the fictional brief and anonymised examples, without personal photographs or confidential documents.

Observable criteria: The prompt describes the subject, composition, atmosphere and exclusions. ; Two results are scored from 0 to 2 for fidelity, coherence, legibility and available space. ; A targeted adjustment is linked to an observed flaw. ; The final choice is justified by the scores and remaining limitations. ; The checklist distinguishes between rights, consent where relevant, notices and approval for distribution..

Frequently asked questions

Do you need to be able to draw?

No. The exercises rely on description and observation, not on drawing skills.

Can you use a photograph of a colleague?

Not during these exercises. Actual use requires checks on consent, rights and internal rules, among other things.

Is the generated image automatically free of rights restrictions?

No. The service's terms and any third-party rights must be examined.

Do you need to disclose that an image was created by AI?

Check the applicable obligations and internal rules; avoid any misleading presentation, particularly for deepfakes.

Last updated: