Key Differences Between FlyWire & Eyewire

Eyewire, FlyWire, citizen science, Rika, Daniela Gamba

If you’re joining FlyWire from our original citizen science project Eyewire, you’ll have swiftly noticed that while interpreting EM slide data remains fundamentally the same between both platforms, many things on FlyWire work differently. These distinctions can be summarized under new commands, new techniques, and new challenges.

Please note that this explanation is currently written with the assumption that you’ve progressed as far as Mystic gameplay in Eyewire. It may be revised over time!

New Commands

FlyWire runs on Google’s web software Neuroglancer, which by default relies on different keyboard and mouse commands than Eyewire’s. The following is not a complete overview of FlyWire controls, but it highlights the most important distinctions from Eyewire’s controls, which you may need some practice to ingrain in your muscle memory:


  • Select Segment: Double left click (not single left click)
  • Deselect Segment: Double left click (not single right click or Ctrl + left click)
  • Hide Paint: Hit the 2 key (not Shift)
  • Decrease Paint Opacity: Open Layer Controls > 2D Visualization > adjust top Opacity slider (not – or +)
  • Undo: Ctrl + Z works as in Eyewire, but it will only undo your cursor movements and segment selections/deselections (it won’t undo merges or splits)
  • Redo: Ctrl + Shift + Z works as in Eyewire, but it will only redo your cursor movements and segment selections/deselections (it won’t redo merges or splits)
  • Zoom: Ctrl + scroll (not Space + Alt + scroll)
  • Pan: Left click + drag, or arrow keys (not Alt + drag or middle click + drag)
  • Up/Down Slice: Only scrolling works for this (as opposed to also using arrow keys or your W and S keys)

2D commands you may remember like Escape to Overview, Explore Mode, Change Brush Size, and Flip Orientation do not exist here. There are no individual cubes to escape; segments are automatically an array of colors; brush size is not adjustable; and you can only view one plane of the data set.


  • Deselect Segment: Double left click (not single right click or Ctrl + left click)
  • Pan: Shift + left click + drag (not right click + drag or middle click + drag)
  • Zoom: Ctrl + scroll (not just scrolling, as this will just move you through 2D slices)
  • Center on Point: Right click (not middle click or Alt + left click)

3D commands you may remember like Jump to Slice and Jump to Neighbor do not exist here, since again there are no individual cubes to work in.

For rotating the 3D cell, left click + drag still works as it does in Eyewire.


FlyWire features a large number of additional hotkeys used for functions that have never existed on Eyewire. Likewise, Eyewire has its own hotkeys for the cell overview and various add-on features; these functions and features do not appear in the Neuroglancer interface. You will find that hitting quite a few keys in FlyWire can produce results you aren’t familiar with at all. For a full rundown of all hotkeys, please see our complete introduction in FlyWire 101, or you can click the three vertical dots in the top right corner of your FlyWire tab, then click “Help.”

New Techniques

In Eyewire, you’ve been helping an AI with limited powers, meaning it will spawn reconstructed volumetric cube-sections based on what’s in parent cubes, starting with a cell’s root cube; and then the AI will run into a problem, which often prevents the cell from growing out all the way. Each cube is “served” to two or more players to decide whether it looks right or not, and once consensus is reached, the advanced players check the cube further, make any additional corrections, and decide the cube is truly complete. The cell is done when an admin takes their own look and doesn’t find anything else wrong. (Of course in Mystics’ case, consensus and full corrections are basically just achieved by two people before an admin sees it.)

FlyWire is both simpler and more complicated than this.

It’s simpler because the AI is more intelligent and robust, giving us essentially a completely reconstructed data set already. What you have to do is look at what’s been classified as an entire cell, decide whether it looks right, and help it look “more” right if you see anything weird. The smallest details generally don’t matter; and once you’re fully trained, you will often be the only person reconstructing your assigned or chosen cell. With practice, you may find yourself completing cells in far less time than it would take a single cell to finish on Eyewire, even with MSTY.

However, FlyWire is more complicated for a few reasons:

  • When you find a missing extension, you must take a step to tell the software that this belongs to the same cell as what you’re already looking at. (See “Merging” in FlyWire 101.) If you just select the missing piece, you’ve only selected it, rather than said anything about a connection.
  • All mergers in FlyWire are fused mergers. That is, when they appear on your cell, they will be marked as the same segment as the rest of the cell. You must choose one of two methods for telling the software that this merger should be separated. (See “Splitting” in FlyWire 101.)
  • These activities rely on getting the hang of your Layer Controls panel, which is unlike anything you’ve needed in Eyewire before. This panel contains many things including your opacity controls, segment color controls, button for marking a cell complete or viewing its completion information, multicut tools (for splitting), and annotation tools (for all kinds of things, really).
  • There are far more cell types to encounter, so you’ll have a learning curve for recognizing a “correct” cell’s morphology.

As you gradually find your way around FlyWire, you’ll probably discover that these differences from Eyewire affect your personal strategies for checking branches and getting mergers down to a manageable size (although at least no mergers can actively run out of control while you’re trying to work!). Feel free to experiment with what works best for you, and please reach out for help if you’re confused!

New Challenges

Some challenges you’ll face on FlyWire derive from the nature of the data set and software. You can view a full explanation of these in our article about [Common Difficulties], but examples include finding branch extensions across bad slides, or re-doing the multicut command due to bad point placement.

Other challenges with FlyWire are more temporary as its citizen science component is still in active testing and feature development. For instance, an equivalent of the Scouts’ Log has not been launched yet, so communication about a cell’s problem areas relies on the chat, the discussion board, e-mail, or the temporary Q&A Log. Similarly, FlyWire doesn’t have in-page notifications. At Boston HQ we remain keenly aware of the Scouts’ Log and notification system’s advantages, so please don’t mistake their absence here for a lack of plans to work on them. In the meantime we greatly appreciate our first alpha testers’ efforts and feedback! Everything you can tell us will assist with building a better FlyWire, and likewise with building the more fully gamified Pyr.

For science!

Leave a Reply

Your email address will not be published. Required fields are marked *