Build a consistent series

How can you preserve the identity of a subject and a graphic world across multiple generated images?

1960 € per session, excl. VAT

In short

In one day of 7 hours, excluding breaks, structure the production of a consistent series of images: select usable references, formalise visual constants, organise trials and correct deviations. You compile a documented production file and assess continuity using a shared assessment grid, within a framework that respects rights.

Why this training?

An attractive image is not enough to make a campaign: altered proportions, fluctuating materials or incompatible lighting weaken product recognition and increase rework. References guide generation without eliminating its variations. Distinguishing identity, style and composition helps you diagnose discrepancies, target corrections and justify choices to the art direction team. Running case study: The fictitious organisation Aveltor Éclairage is preparing six visuals of an imaginary desk lamp for a catalogue: three settings, with two framings each. The silhouette, proportions and finish must remain stable despite changes in the surroundings. Transfer to work: For your next series, have the constants approved before generation, test two contrasting images and then expand production. Compare the number of critical discrepancies between versions and archive approval decisions.

Objectives of this training

  • Classify references according to their role and conditions of use.
  • Formalise constants for identity, material and light.
  • Produce six variants following a documented trial protocol.
  • Measure continuity discrepancies using an explicit assessment grid.
  • Correct priority deviations and prepare the file for handover.

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Graphic designers and art directors producing AI-assisted series.
  • Visual content managers and communications teams overseeing variants.

Prior skills

  • Have prior experience of generating images and adjusting a descriptive prompt.
  • Be able to analyse framing, palette, proportions and lighting.
  • Have access to a tool authorised by your organisation.

What you will work on

  1. Define the expected continuity

    Objective: Separate constants from permissible variations. Product identity, proportions, material, palette; the difference between stylistic consistency and fidelity to the subject.

  2. Assess the references

    Objective: Compile a usable reference set without contradictions. Identity, style and composition references; rights, individuals’ consent, confidentiality, notices and risks of deceptive deepfakes.

  3. Establish a benchmark image

    Objective: Translate constants into verifiable instructions. Hierarchically structured description, contradictions between references, available settings; limits to reproducibility depending on the tool.

  4. Create variants without losing identity

    Objective: Build the series while controlling variations. Stable descriptive block, variables for surroundings and framing; use of references according to the functions actually available.

  5. Assess and correct deviations

    Objective: Prioritise rework based on observable discrepancies. Side-by-side comparison; ratings of compliant, minor discrepancy or critical discrepancy for silhouette, proportions, finish and light.

  6. Prepare for editorial approval

    Objective: Hand over a traceable series together with its limitations. Naming, selection, distribution conditions, appropriate notices and human approval; distinction between conceptual illustration and faithful representation.

What you take away

  • Reference library with a rights register.
  • Constants sheet and matrix of the six variants.
  • Image series with a log of trials and corrections.
  • Completed continuity assessment grid and handover file.

Method and assessment

  • Short teaching inputs followed by immediate practice on the running case study.
  • Individual work followed by work in pairs, with peer review of the outputs.
  • Guided group discussion: what works, what needs correcting and what remains a human decision.
  • Preparation: Prepare a fictitious or anonymised brief and references with documented usage rights. Check the available functions, confidentiality rules and export options of the tool authorised by your organisation.

Observable criteria: Each reference has an explicit provenance, role and usage status. ; The six images are compared against the same constants. ; Each critical discrepancy is corrected or flagged as a blocking issue. ; The selected versions are linked to the prompts and settings actually used. ; Consent and notice requirements are examined before distribution..

Frequently asked questions

Is one reference enough to preserve a product exactly?

No. Fidelity is checked image by image; some discrepancies require retouching or another production method.

Can you use a portrait of a real person?

Only with the necessary permissions and appropriate consent; the exercises favour fictitious subjects.

What should you do without a visual reference function?

Use a structured description and manual comparison, documenting the limitations encountered.

Must the use of AI be disclosed?

Examine the applicable obligations and editorial rules, particularly to avoid any confusion or misleading presentation.

Last updated: