A deep dive into Consistency in Media Production with AI

Facilitator: Markus Tischner, Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany.

Getting started with AI-supported media production often brings familiar challenges: translating an idea into a useful prompt, generating the right scene, and keeping a character recognisable across several images. A setting changes unexpectedly, the visual style drifts, or details disappear when an image becomes a video. Consistency quickly becomes a fundamental production challenge. This practical workshop explores how to address these challenges through prompting, co-creation and workflow design. We examine how dialogue with AI can help clarify an idea and develop effective prompts, and how reference images, character sheets and visual briefs can guide a coherent production.

Practical examples and live demonstrations introduce the approaches in the first two parts. Participants then create their own media, followed by a shared review that makes room for discussing results, difficulties and production decisions.

Workshop leader Markus Tischner is a filmmaker and learning expert who heads media production at the Competence Centre Education at Friedrich-Alexander-Universität Erlangen-Nürnberg. Since 2022, he has been exploring generative AI in learning design and media production. He brings practical experience with character development, image series, video generation and workflows combining AI-generated material with stock media and original footage.

The three-hour workshop

Part 1: From an initial idea to a consistent visual foundation: approximately 35 minutes

Through practical examples and live demonstrations, we explore typical difficulties when beginning AI-supported media production. We examine how to develop prompts in dialogue with AI: asking questions, explaining an idea, inviting clarifying questions and working together on a suitable generation prompt. We then identify what needs to remain consistent across a production: character identity, appearance, recurring locations and objects, and visual style. Examples show how visual briefs, character sheets and reference images can turn these requirements into a usable foundation for generating related assets.

Part 2: From text to image to video: approximately 45 minutes

Practical examples and live demonstrations follow the development of related images and their transition into moving sequences. We establish a visual reference, generate variations in pose, viewpoint and scene, and examine how targeted revisions can help preserve continuity. We then explore image-to-video generation, assessing what happens to characters, settings and visual details when movement is introduced. Throughout, we discuss when to refine a prompt, revise an image, generate another variation or assemble existing material.

15-minute break

Part 3: Create, share and review: approximately 85 minutes

Creation phase: approximately 45 minutes

Working in small groups, participants develop two or three related images for an educational use case. You use AI dialogue to clarify your idea, define the features that need to remain consistent, and develop prompts and visual references. You test and revise the resulting image sequence and consider how a selected image could become a video shot.

Presentation and shared review: approximately 40 minutes

Each group shares its image sequence and briefly explains its approach. Together, we examine what remained consistent, where unwanted changes appeared, and how prompts, references and revisions influenced the results. The discussion focuses on production decisions and possible improvements: What would you change in the next iteration? Which approach from another group could help your own work? Comparing the sequences allows participants to learn from a range of attempts and identify useful principles for their own workflows.

What you will take away

• Practical approaches to developing prompts through co-creation with AI.
• An understanding of character, scene and style consistency.
• Ways to use visual briefs, character sheets and reference images.
• A workflow connecting text-to-image and image-to-video generation.
• An understanding of how AI reshapes traditional production phases.
• A first consistency workflow, informed by your own experiments and shared feedback.

Who should join

Media producers, video teams, instructional designers and learning technologists who use, or are beginning to use, generative AI for educational media. Basic experience with AI image generation is helpful. Programming skills are not required.

Before the workshop

Bring a small production idea involving a recurring character, setting or visual style, such as an illustrated learning scenario or a short explanatory sequence. If available, bring an existing image or visual reference to use as a starting point.

Due to the workshop format, places are limited to 30 participants. Register early to secure your spot.

The workshop is free for MLA members and costs €75 for non-members.