Background
Meaning Machine is a Bristol-based game developer and the creator of an upcoming murder mystery video game, Dead Meat, in which players assume the role of a detective interrogating suspects. The technological twist on this setup is that players can ask game characters anything they like and the characters will answer back. The game uses generative AI to shape dialogue in response to the player’s questions, and players have the option to interact through voice or keyboard input.
Screenshot from Dead Meat, copyright Meaning Machine.
Application of AI
Dead Meat’s responsive dialogue is generated by language models. Early versions of the game used cloud-based large language models (LLM), but cloud-based LLMs introduce cost and latency issues, as well requiring players to have an always-on internet connection. Meaning Machine then tried fine-tuning a smaller language model, that could run on device, in partnership with Nvidia. The model was tuned on Meaning Machine’s own dialogue data and created good results but the output was still sometimes inconsistent or unpredictable, getting stuck in narrative loops with no satisfying storytelling beats.
The team have since worked on a proof-of-concept, Blood Will Out, in collaboration with Nvidia and the University of Bristol. Through this they have implemented what studio co-founder and creative lead Thomas Keane calls, with tongue firmly in cheek, “the bully system”: a “hyper-directive, author-based control over what AI can do at the point of generation”. Keane says this avoids the “spray and pray, ‘I hope you can do what we do’ kind of vibe” when using generative AI in creative contexts.
Meaning Machine hope to be able to share more news on Dead Meat in the near future, and the research from University of Bristol will be available in a forthcoming paper.
“We are pretty optimistic about AI… not as a productivity tool, not as something to build things faster and cheaper. We are optimistic about AI as something that can unlock new forms of creative self-expression for developers and create new types of experience for players that radically change the shape of what video games look like.”
Thomas Keane, co-founder and creative lead, Meaning Machine speaking at Game Developers Conference, May 1, 2025.
Applying the CoSTAR Foresight Lab AI roadmap
Our AI roadmap is organised around three strategic outcomes – frameworks, targeted support, and growth – and driven by nine recommendations that seek to align technological advancement with ethical responsibility and economic opportunity, ensuring long-term growth and success of the UK screen sector.
How this case study aligns with the roadmap
- Carbon
- Switching from using large language models in the cloud to a small language model that runs on-device can be a positive step towards carbon minimisation, as it avoids the cost and energy use associated with LLM inference and cloud compute.
- Responsible AI
- Meaning Machine’s partnership with Nvidia demonstrates the sort of cross-discipline collaboration that will help deliver responsible AI for the screen sector. The partnership between AI developer and game designer has resulted in an AI solution tailored to the needs of a specific game experience.
- Investment
- Meaning Machine’s AI innovations have been supported by funding from the MyWorld programme, illustrating one way in which investment drives the emergence of new creative technologies and new creative technology businesses.
- Independent creation
- Dead Meat is an example of UK creators leveraging AI technologies to pursue new storytelling ideas, formats and concepts.
Resources
Citation
@online{tarran2026,
author = {Tarran, Brian and Johnston, David},
publisher = {CoSTAR Foresight Lab},
title = {Meaning {Machine:} {Fine-tuning} Language Models for
Responsive Character Dialogue},
date = {2026-03-03},
langid = {en}
}