In order to make use of the Auto Caption feature, you need to have a Caption Config enabled. There are a number of Config files that can be used by everyone, but you are also able to edit these and create your own.
A Config file contains settings which will determine the output of the Auto Caption feature.
People Formula Configuration
The people formula configuration determines how the names of the people will appear in your caption. This is where you can set whether you want a players position, number or team to appear in the caption as well. These settings are the same as the settings in a normal =people formula, and you can read an in-depth article about those here.
Caption List
The caption list contains two columns, the "Value" and the "LLM Statement".
Value Column
The Value column must contain the full caption structure including all desired formulas and options. This determines what the software outputs if this caption is selected.
Formulas
Below is a list of all available formulas, along with a description of what they do and how they are used in the Config:
Formula | Notes | Example |
=person() | This formula will output a single person's name | =person() throws the ball |
=people() | This formula will output any number of people that are detected in the image (excluding anyone already appearing in a =person() formula) | =people() battle for possession |
=plural("singular value", "plural value") | This formula will output one of two options depending on whether a single or multiple people are present.
1 person = first value 0 or 2+ people = second value | =plural("celebrates", "celebrate") |
=nth() | Outputs a list option to be used to select between numbers, specifically: first, second, third, etc.. | during the =nth() inning |
=team() | This formula is ideal for sports where you have multiple active groups in the project. It will output the OPPOSING group name (team name) to the person in the caption | =person() scores against =team() |
=prepend(prefix, formula) | This formula adds text before the output of a different specified formula, and it will only output content if the included formula has content.
The second value in this formula MUST be a formula. The prefix can be text or another formula. | =person() runs the ball =prepend("whilst under pressure from", =person())
In this example, if only one person is detected in the image it will say:
"Kristian Fulton (8) runs the ball"
However, if a second person is detected in the image it will say: |
=fallback(primary, backup) | This formula will output a backup value if the primary value produces and empty result. The primary value MUST be a formula, the secondary value can be either text or a formula. | =person() stiff arms =fallback(person(), "opposing defender")
In this example, if only one person is detected it will output:
"Kristian Fulton (8) stiff arms opposing defender"
However if the second person is identified, it will change to:
"Kristian Fulton (8) stiff arms Dont'e Thornton Jr. (10)" |
=variable("ID", "option 1", "option 2", "option 3") | This formula creates a user-selectable dropdown with predefined options.
The first value in the list will be the unique identifier of the list, the subsequent entries will be the list options.
Leaving an entry empty by adding "" will present an "empty" option.
| =person() celebrates =variable("celebrations","", "an interception", "a first down", "a touchdown", "after scoring", "the win", "a deflected pass")
In this example, an interactive list will be presented to the user so that they can choose what the person in the image is celebrating. See the screenshot below.
|
Variable Formula Example
LLM Statement Column
The LLM statement column in the config file is essentially what the AI is reading when it determines the relevant caption. By default, this is just a simplified translation of the Value field but you can edit this manually and provide additional information to help the AI make its decision.
Here are a few examples of when this might be useful:
In the following two examples, the caption is only really relevant if the person in the image is the quarterback, so we have specified in the LLM column that it is specifically the quarterback being sacked or preparing to take the snap:
In the following example, we have an action of "punts the ball". Sometimes the AI can be confused between a punt and other types of kicks, so here we have specified exactly what we are looking for when we ask for a punt.
In this example, the only person who we want to caption giving the team instructions is the manager, so we have said in our LLM statement that this caption is the manager giving instructions.
In this example, we want to make sure the bus arrivals are considered when assessing if a player is arriving at the stadium prior to a game, so we have included this as an additional instruction in the LLM statement.
In all of these examples, the software will still only output the corresponding Value field. The LLM statement field is just what is presented to the AI in order for it to make its selection.
Editing Configs
If you would like to edit a Config file or just explore the different use cases of the above formulas, you can select a global config from the project settings, select "save as my copy" and then you can "edit config".