Custom Metadata

Start with what you want metadata to help someone do: find the right footage, filter a library, or understand an asset during review. That tells you which information to keep and how to add it. There are four useful approaches: manual fields, AI-generated fields, people recognition, and object recognition. You can combine them in the same project. For example, let Aspect recognize who’s in a clip while your team sets its approval status.

1. Decide whether you need a field

Write down the question you want to answer before creating anything. A field is useful when you need a repeatable label, a value your team controls, or information that should stay visible alongside an asset. A one-off visual search often needs none of that. Technical details such as duration, resolution, and frame rate are already extracted automatically. Check the asset’s General tab before making a custom field for information Aspect already records.

2. Choose how the values will be added

Manual fields: information your team knows or decides

Use manual fields for campaign names, usage rights, approval status, or anything you want a person to enter deliberately. For example, create a Talent field with a Multi Select option for each person’s name, then select the people who appear in each clip. Multi Select lets one clip contain several people; Single Select is better for a single status such as In review or Approved. Open Settings → your project → Metadata → Create attribute, choose a name and data type, and leave Generate with AI off. You can enter values on an asset, edit them in List view, or apply them to multiple selected assets. Use this approach when a small set of clips is easy to label, or when the value reflects a human decision. AI can describe a shot, but it cannot know that your client has approved it unless someone supplies that information. See field types and editing values. For an example of manual fields in use, watch how a team tracks asset status and assigns editors.

AI-generated fields: your categories, applied from content

Use AI generation when you want the same question answered across many assets. It fills a custom field using Visual content, the Transcript, or both. For example, create a Shot type field with Single Select options: Interview, Product close-up, and B-roll. Turn on Generate with AI, choose Visual, and write guidance such as:

Choose the main shot type: Interview for a person speaking to camera; Product close-up for a detailed product shot; B-roll for supporting footage. Leave it empty if none applies.

A transcript-based example would be an Interview summary text field that summarizes the main topic discussed in each clip. Choose options that your team will actually filter by, and test the guidance on a small folder before applying it broadly. Review the results and adjust the prompt when categories overlap or values are too vague. See AI-generated values.

People recognition: find a particular person

To find a particular person, you can either use the manual Talent field above or teach Aspect to recognize them from reference photos. Open Settings → your project → Metadata and choose Create person under People Recognition. Enter the person’s name, add clear reference photos, and crop each photo to their face. Once the references have finished processing, run recognition on the relevant footage. Aspect records matches in the project’s People field. You can then filter by that person or use their name in a search. You can also correct whether they appear on a particular asset yourself. Use recognition when the same person appears across a growing library. Use manual labeling when you already know the clips and want to assign the labels yourself. See people recognition.

Object recognition: identify something specific

Use object recognition for a particular product, vehicle, prop, or other object that matters to your team. Generic objects such as headphones are usually a good starting point for Visual search; your specific headphone model may deserve an enrolled object. Choose Create object under Object Recognition, give it a name and description, and add reference photos. Configure the Recognition policy to explain which details can vary and which must match. For example, a product’s color may vary while its identifying label must stay the same. Aspect records matches in the project’s Objects field, which you can use in filters and search. See object recognition.

3. Run it on existing footage, then decide what to watch

For an AI-generated field, person, or object, start with a folder whose results you can check:

1. Open the folder in List view and show the relevant metadata column.
2. Open the column’s header menu. For People or Objects, choose the person or object to run.
3. Choose Run remaining for the current folder to process assets that haven’t been answered yet. Use Re-run all when you want to revisit earlier AI results.
4. Review the results, correct mistakes, and refine the prompt or reference photos if needed.

Turn on Watch in the field, person, or object settings when you want it to run automatically on new assets added to selected folders. Choose the folders and asset types it should cover. Watch does not backfill existing footage. Creating or saving the setup is separate from running it on existing assets. See run scopes and re-runs for the available controls.

4. Put the metadata to work

Use Display to show the fields your team needs in the library, then Add filter to narrow the assets. You might filter People to a specific interviewee and Status to Approved before searching for a particular line from their interview. For reviewers, decide which information helps them do their job. A client may need Status, Description, and Usage instructions, while Internal notes should stay hidden. In the share-link settings, enable metadata visibility and choose the visible attributes. Metadata editing is a separate permission. If you need to create fields or update many assets at once, Aspect Agent can help. Give it the assets, field, and intended values, and review the proposed changes. Start with a few useful fields. Add more when you can explain who will use them and what they will help that person find, filter, or understand.

Make the fields useful in your library

Watch how to show metadata, filter unwanted file types, and sort the library into a useful working view.