Media creation with artificial intelligence

Copyright and fair use

To understand the full picture of copyright, it is necessary to consider its real-world implementation.

Game companies have a strong interest in not upsetting their fan base, especially organized structures such as gaming communities.

Game communities function as voluntary support structures that provide services companies would otherwise need to fund themselves. Many of these contributions can be understood as direct or indirect sales promotion. These communities typically offer:

• First contact and general help for newbies.
• The creation of guides, tips and tricks, and even complete playthroughs.
• Support in organizing competitions and other promotional events, including creative activities such as fan art. Cosplay contributions can increase a company's visibility at events like Gamescom.
• An additional channel for updates (information and content).
• Bug reporting and, in some cases, bug fixing. In rare instances, community members may even contribute to maintaining source code.
• Mods that improve replay value and thus increase overall customer satisfaction.
• Increased likelihood that fans will purchase other games and products (merchandise) from the company.
• A pool of trusted and engaged players who can be recruited as beta testers for new releases.

As a result, most companies employ community managers. In practice, companies often tolerate limited uses of their intellectual property because they benefit from these activities. A strictly enforced copyright regime could suppress creative community contributions, reduce engagement, and ultimately harm the company itself.

However, this does not mean that copyright is overridden. In some cases, fan works may fall under fair use (depending on jurisdiction), but more often they exist within a space of informal tolerance or explicit licensing policies.

Modifications (mods) often remain short of becoming independent games:

• Typically, they add optional 2D and 3D content, sometimes in large quantities. However, more fundamental changes - such as new game mechanics - often require access to or modification of the game engine and are therefore limited.

Generative AI (GenAI) has the potential to disrupt this cost–benefit balance:

• The large-scale or automated production of new content based on existing assets may conflict with the company’s interests.

Considering the artists and technicians involved in creating the original work, mass-produced fan content generated at little or no cost could threaten established creative professions.

• An overabundance of derivative content may dilute attention and reduce consumer motivation to engage with official products.

The goal should be a form of symbiotic coexistence:

• Game communities should avoid creating direct competition with official products.
• In contrast to official expansion packs and DLCs, mods should generally not be placed behind paywalls.

As of 2026, conflicts arising from GenAI-driven mods remain largely hypothetical, but their relevance is likely to increase. In the long term, cooperative development models between companies and communities are conceivable, though this remains uncharted territory.

Possibilities and limitations

Generative artificial intelligence (GenAI) can ease and accelerate content creation. The difficulty of using or creating your own setups will continue to decrease with each newly released commercial forerunner model.

You can use GenAI via websites, desktop clients, or dedicated programs that may run fully locally. To build your own programmatic solutions, you will either need downloadable AI models or API keys to access cloud services that perform the heavy computation remotely. With sufficient expertise, you can even build agentic AIs (such as Open Claw) that use your existing tools and carry out tasks automatically. However, caution is advised: probabilistic AIs can hallucinate and may pose a risk to your system. As a mitigation, MCPs should be used. Sandboxes offer an additional layer of safety, but they can limit the usefulness of agents and introduce extra complexity, which may offset the time savings you intended to achieve.

GenAI systems operate probabilistically. Do not expect identical results when repeating prompts with the same inputs. The same text prompts may produce similar, but not identical, outputs. Therefore, in some scenarios, it can be beneficial to generate multiple results and select the most suitable candidate for your intermediate or final goal.

Sounds, voice acting and music

• Voice-cloning of existing or creation of new voices. For natural voices you may want to look for emotional text-to-speech.
• Music generations

Image generation

• Content generation based on:
  • Text prompts
  • Own drafts
  • Merging (main image and references)
• Changing existing content
  • Expanding
  • Inpainting (replacement of subsections)
  • Style transfers

3D content generation

There exists content generators that turn 2D data into 3D data by calculating plausible assumptions for the missing dimension.

Tools

2D images

All big LLM applications such as ChatGPT, Claude, Gemini, Grok and Mistral support image generation. Of course there are other tools but you probably already at least one of these. Therefore you have there an account and can instantly use it for image generation. For mass production you probably need a paid subscription or "plan" so that more images can be created in a defined time frame (by that plan).

For image editing you also want a specialized graphics tool such a gimp, krita or Photoshop (which itself has GenAI functions).

Easy Access: ChatGPT (The notes here may work the same for other well known LLMs.)

• Prompt exactly what you want (even if it just "higher quality"). Either it works or not.
• Merging: When possible drop both images at the same time into the prompt. Re-editing an image means a loss of details.
• The the context window gained to much control over the currently expected output, then start a new prompt that includes all the accumulated changes you want to make.
• When you subscription allows it, output multiple final images. As every piece will be different, use have to chose the best one. You can also photoshop (merge) multiple final images to together by using masks.

(Add some examples here.)

Specialized: Canva

Local solutions: AUTOMATIC1111 (aka Stable diffusion), ComfyUI

Videos

Grok (xAI)

Gemini (Google)

Sora (OpenAI)

3D objects

• ...

3D animations

• ...

World generators

• World generator inside Unreal Engine 5
• ...