Overview
A Moment type defines the segment of signal that you want to identify and the information you want returned. A moment type tells the AI which kind of moment to spot in your streams or files (a try in rugby, a breaking news item…).
To manage your collection of moment types, open the Moment Manager page from the AI icon in the left hand menu bar. Moment types are grouped by category. Use the toggle at top right to switch between grid and list views; the list also shows each type's detection Signals and Source.
Moment types page in grid view
Moment types page in list view
Create a moment type
Decide first which category the new type belongs to. You can then create it from scratch or from an existing type.
From scratch
- Click the AI icon in the left menu bar, then + Create.
- Select Moment type (the other choice is Category).
- Fill in the empty New moment type window, using the settings below, then click Create.
From an existing type
- Click the … button at the end of a moment type row (list view) or top right corner of a moment type card (grid view).
- Select Duplicate. A copy named "(copy)" appears in the same category.
- Click the copy, or … > Edit, to open the Edit moment type window.
Moment type ... menu: edit, duplicate, delete
In the Definition area:
- Category: pre-filled with the source category. Pick another one, or click Add category to create it.
- Template: reuses the definition of any existing moment type, even from another category.
- Name: this is the moment type name, make it descriptive. Use names that operators can understand immediately, such as Interview, Weather Report, or Post-match Interview..
Set up detection
Describe what qualifies as a match. Where signal selection is available, review the inputs used for detection. Signals describe the information being analyzed; sources describe where analysis runs.
Detection signals: define the technology used for detecting the moment type. Click the signals that fit your moment type.
- Transcript: AI speech-to-text of the source.
- SCTE: SCTE-35 broadcast markers (programme or chapter start/end, ad-break boundaries, custom markers).
- Scene analysis: audio, video and OCR analysis.
Instruction: the prompt that tells the AI what to look for. Useful instructions explain: (1) what must be present; (2) where the segment begins and ends; (3) what should be excluded; (4) how to distinguish it from similar content.
We recommend to structure your instruction prompt with these four parts:
- Identify: what typically happens ("the referee shows a yellow or red card to a player").
- What counts: the elements of interest (the offence, the referee's decision, the player leaving).
- Boundaries: precisely where the moment starts and ends, with any special cases.
- Not this: exceptions that must not be detected (a warning without a card).
Definition and detection settings
Processing step (seconds): length of the chunks analyzed. With 300, the source is scanned in 5-minute chunks.
Source: where detection runs. Without any source, it runs nowhere.
- Selected streams: pick one or more streams.
- Source tags: any stream carrying one of these tags is covered, for example rugby.
- Import files: when on, imported files with the source tag are also analyzed.
Processing steps and source settings
Metadata
Review the fields that should be generated for each detected Moment.
A field definition can specify its name, data type, required or optional status, default value, and AI filling instruction. Available controls depend on the deployed interface.
Example: Weather Report
Category: News
Type name: Weather Report
Suggested detection instruction:
“Identify dedicated weather forecast segments. Include the presenter’s introduction and forecast explanation. Exclude brief weather references within other news stories and promotional teasers.”
Suggested metadata:
| Field | Suggested AI Filling Instruction |
|---|---|
| Presenter | Identify the presenter when supported by the available content. |
| Region | Identify the geographic area covered by the forecast. |
| Dominant condition | Summarize the main forecast condition described in the segment. |
These are example settings, not guaranteed generated values.
Click Save. The moment type is now created.
Edit or delete a moment type
Use the … menu on any row: Edit reopens the settings above, Duplicate copies the type, Delete removes it.
Category settings also apply: streams and tags set on a category are inherited by its moment types. If you change a type's source, the window warns which category source it removes. See Managing AI Moment Categories.
Built-In Types
Built-in definitions are maintained by Wildmoka. Regular configuration users should not assume they can rewrite those definitions, although their source scope can be edited.
FAQ
What is the difference between a category and a type?
A category groups related types and provides shared context. A type defines a particular detection target and its metadata.
Why does a type apply to an unexpected source?
Review the type’s source configuration, inherited category scopes, and applicable source tags.
Why does a duplicated type have different sources?
The moment type duplication behavior inherits the category’s scope rather than copying the original type’s scope. Review it before use.
Can I define my own editorial Moments?
Yes. Custom types can express your detection criteria and metadata requirements—for example, interviews, weather reports, or studio segments.
Can I change a built-in type?
You can configure its source scope where permitted. Built-in detection definitions are maintained by Wildmoka and have more restricted editing access.