A hands-free mouse grid by Ben Rollin, Tara Roys, timotimo, and aegis, for use with the Talon voice framework. Flex Mouse Grid is just the latest in a series of iterations, and was built on top of Dense Mouse Grid.

You can think of Flex Mouse Grid as extending the functionality of Dense Mouse Grid, attempting to be an even more flexible, all-in-one voice mouse tool. Some notable differences from Dense Mouse Grid:

• "Points" mapping. Essentially incorporating Andrew Dant's awesome screen-spots tool and adding a light UI. It allows points to be saved on an application-specific basis with a custom name. Example: point click continue to click a saved point called "continue".
• "Box" detection. Use image processing techniques to detect boxes on the screen, identifying them and allowing you to click them. You can also save boxes as points. Example: boxes to identify all boxes in the active window, and then box click seven to click the box labeled "7".

1. Ensure you have Talon installed.
2. If you haven't already, clone the talonhub/community (formerly knausj_talon) repository into your Talon user directory. This tool is designed to be used along with the community repository, but is compatible with any repository that provides the letters a-z using the <user.letters> capture and numbers using the <numbers> capture.
3. Install the python dependencies with the below instructions.

• On Mac or Linux, run these commands in a terminal:

~/.talon/bin/pip install opencv-python-headless
~/.talon/bin/pip install numpy

• On Windows, run the following commands in a Powershell. Note: on Windows the bin folder is a shortcut file that cannot be simply navigated to.

cd (New-Object -ComObject WScript.Shell).CreateShortcut("$env:USERPROFILE/AppData/Roaming/talon/bin.lnk").TargetPath
./pip.bat install opencv-python-headless
./pip.bat install numpy

4. Clone this repository into your Talon user directory (on MacOS, located at ~/.talon/user).
5. Restart Talon, can't hurt.
6. Try the voice command flex grid!

If you are just interested in the dense mouse grid features, you may consider just using the simpler Dense Mouse Grid. Otherwise, you can read the following sections on Points, Boxes, and finally Dense Mouse Grid to learn about the core features of this tool.

Note that some of the voice commands for the Dense Mouse Grid have been renamed. Mostly swapping the order of certain commands like going from smaller grid to grid smaller. Feel free to change these back of course!

Flex mouse grid allows saving "points" which are just labeled points on the screen. They are always saved relative to the active window, so you can safely move a window around, and know that points will be unaffected.

• map <word> creates a new point at the current location of the mouse cursor with the label "word"
• unmap <word> deletes the point labeled "word"
• unmap everything deletes all points for the active window. Careful with this one!

• points shows all points for the current window
• points close hides all points
• grid close hides all points, all boxes, and the dense grid

• point <word> moves to the point labeled "word"
• point click <word> clicks the point labeled "word"
• point righty <word> right clicks the point labeled "word"

(Relevant when mapping boxes as points) This is a list of points all with the same label. You can move to/click the third point in a point list simply with

• point <word> three

Note that the list is 1-indexed. If you omit the index, it is assumed to be one. Other clicking commands also work in a similar way to the above.

TODO

For inaccessible programs such as some games where there are visually distinguishable boxes which you would like a quick way to click, there is the boxes feature of flex mouse grid. It attempts to identify boxes with computer vision/image processing, and labels the boxes that it found, allowing you to interact with them.

Lets take as an example this set of boxes from a game:

• boxes finds boxes labeling each with a number. It attempts to find the optimal box detection configuration for the current window

• box <number> moves the cursor to the specified box, hiding the boxes afterwards
• box click <number> clicks the specified box, hiding the boxes afterwards
• box righty <number> right clicks the specified box, hiding the boxes afterwards

• boxes close hides all boxes
• grid close hides all boxes, all points, and the dense grid

Make sure to learn about Points first.

• map <word> box <number> creates a new point at the center of the chosen box labeled "word"
• map <word> box <number1> mark <number2>... creates a point list of two points. You can include more points by including more mark <number>s.
• map <word> box <number1> past <number2> creates a point list that will include a point for every box in the range between "box number1" and "box number2". Can also be done in reverse.

So if we said boxes and saw:

We could say map sprint box five to create a point there:

Then we can simply move to the "sprint" button with the phrase point sprint

Further, maybe we just want to refer to all of these boxes as actions. We could say a command like: map act box thirteen past three which would create a point list as seen below:

It may be a little hard to tell what's going on, but each box got put into a point list with the label "act". We could then move our cursor to the same "sprint" button with the following: point act nine.

Box detection works by doing image processing with three configurable parameters. By default the boxes command will attempt to set an optimal threshold parameter, but depending on the color scheme of the window, you may be able to improve it. The threshold, upper bound, and lower bound can be configured via voice so that more boxes can be found.

There are three box configuration parameters that are currently changeable:

• upper box size upper bound. Boxes will not be larger than this number of pixels.
• lower box size lower bound. Boxes will not be smaller than this number of pixels.
• threshold the filtering threshold. The higher the number, the lighter that colors can be distinguished.

To see what parameters are already set:

• flex info toggles a view of the current grid and box configuration parameters

You can modify these parameters on an application specific basis with the following commands:

• boxes <parameter> more
• boxes <parameter> less

So e.g. boxes upper more would increase the maximum size that a box can be. After changing any parameter, boxes will be found again so you can gauge if it was a good change or not in real time. You can additionally modify parameters by a smaller amount with bump:

• boxes <parameter> more bump
• boxes <parameter> less bump

Hopefully the upper and lower parameters make intuitive sense, but threshold is probably a little more mysterious. The best way to understand it is to see what it is doing. You can use the command:

• boxes threshold

to toggle an overlay of the processed image. This image will be black and white, and based on the threshold parameter setting, different parts of the image will be filtered out. Once you have turned on the overlay, you can begin adjusting the threshold with e.g. boxes threshold more and see the resulting change. This can be a useful way to hone in on the best threshold.

The Dense Mouse Grid is a completely hands-free replacement for traditional mousing. It replaces mouse movements with voice commands. You tell the computer what grid coordinate you want to move to on the screen, and the mouse grid moves your cursor to that location.

This is the "dense mouse grid". It fills the screen with fields that can be reached with a number and two letters.

• flex grid to show the grid. It is displayed over the active window by default
• flex grid screen to put the grid over the whole screen
• flex grid screen <number> to put the grid over a different screen
• grid close to close the grid

The grid will appear in Frame Mode by default. Say:

• cap bat

This will move the cursor to row "c" and column "b". You can include a number to move the cursor to another block (3 drum wax). To click, use whatever voice command you have associated with clicking. By default in knausj_talon, this is touch.

If you have selected the wrong number, you can choose a different number anytime. If you have selected the wrong row, you can say row <letter>, and it will move the highlighed row to that letter. So saying

• one
• harp
• row bat
• cap

will result in moving the cursor to coordinate 1 B C.

If you want to type a letter or number while the great is open, you can say "press" followed by the letter or number.

• press air types "a"
• press bat cap two types "bc2"

If frame mode does not suit you, there are two other modes that can be used to show the coordinates.

• grid checker turns on checker mode, which overlays the coordinates in a checker pattern. This can be visually confusing, but it splits the difference between being able to immediately see the coordinate and being able to see the contents of your screen.
• grid full shows every possible coordinate.
• grid phonetic switch to phonetic mode, which is just like frame mode except with full phonetic words labeling the rows and columns instead of individual letters.
• grid frame switches to frame mode (the default).

• grid bigger
• grid smaller

You can adjust the size by a smaller amount by including bump, e.g. grid bigger bump.

The default size of the grid can be set in flex_mouse_settings.talon by changing the number in user.flex_mouse_grid_field_size = "30".

The grid can be made more or less transparent with the following commands.

• background lighter
• background darker
• letters lighter
• letters darker

You can also include the word bump to adjust the value by a smaller amount. For instance, background darker bump.

Every color in the grid is modifiable in flex_mouse_settings.talon, allowing you to set the defaults to whatever is comfortable.

The colors use 6 digit hexadecimal RGB colors. The transparency uses 2-digit hexadecimal numbers for an alpha channel.

• Configurable characters: alphabet, alphabet subset, numeric
• Deduplicate grid config state
• Allow clearing saved grid config, box config
• Include more things in grid config
• Pipe image data to new python process as binary