Week 6

Week 6

This week we did some more design iteration on our 3 ideas.


Picking up from last week, we started building our puzzle prototype this week. Our goals with this prototype are:

  • To test if people enjoy solving an obstacle based puzzle by exploring and choosing animations
  • Create and test a user interface which players use to explore and choose animations

We tweaked the design of our puzzle prototype and came up with a simple puzzle level. Here’s a short version of the design of the puzzle prototype:


  1. Show the goal (fish) and map of the level
  2. The player uploads videos of movements that they think will be required to reach the goal

GOAL : Eat fish


  1. River
  2. Tower which holds fish


  1. Players must upload animations to interact with boxes
  2. Use any movement whenever needed


  1. Cat cannot swim in the water
  2. Boxes stay in place inside the water

Apart from this we also tweaked the design of the fighting idea a little.


Our artist created art assets for the puzzle level. Here are some snapshots

Our artist also created an enemy character for the fighting prototype.


Our gameplay programmer is currently building the puzzle prototype and we are planning to have a working demo of this next week. We plan to playtest this prototype and also get feedback from our client next week. Currently we have a character in Unreal whose movement is physics based.

We are also working to get all the different components of the machine learning pipeline working. The machine learning pipeline is made up of 3 parts: Human Mesh Recovery (HMR), Motion Reconstruction and DeepMimic. For reference you can find the breakdown of the technology in this technical blog post (coming soon).

In our meeting with our client last week we decided that we will not be using the third part of the pipeline, DeepMimic, since its function can be achieved by using a game engine like Unreal Engine.

Next Week

In the next week we will continue to work on the pipeline (from 2D video to 3D animations in Unreal). We want to record and test videos of different types of human movements to check which movements work well with the algorithm and which don’t. We want to finish implementing the puzzle prototype next week. We also want to come up with a concrete design for our storytelling/party game prototype.