AI Enrichments
AI enrichments are processing steps that run automatically on new audio and image content. They can transcribe a recording, fill in your own fields from what was said, and caption and tag photos, so your media becomes searchable and structured without anyone typing it up.
AI enrichments may need to be enabled for your organisation. Each one (AI Transcription, AI Extraction, AI Vision and AI Tagging) is switched on separately, along with Qik's AI master switch. If you don't see the options described here, contact support.
The available enrichments
| Enrichment | Works on | What it produces |
|---|---|---|
| Transcribe Audio | Audio content | A text transcript of the recording, with timings. Optionally labels the different speakers. |
| Extract Field Data | Content that has been transcribed | Values for the content type's own fields, worked out from the transcript. |
| Describe Image | Images | A one-sentence caption describing the scene, plus 5 to 12 concept tags (people, activity, setting and mood). |
Separately, AI tagging of uploaded images doesn't need any setup on a content type. When it is enabled for your organisation, every image you upload is automatically labelled with the objects it contains (for example "Person", "Bicycle", "Stage").
How each enrichment helps
Transcribe Audio
Record a meeting, an interview or a voice note, upload it as audio, and the transcript appears on the record once processing finishes. You can set the Language (a code such as en or en-AU; English is used if left blank) and turn on Diarize to label who is speaking. Speaker labels take a little longer.
Extract Field Data
Extraction reads the transcript and fills in the fields you've added to the content type. For example, a "Site Visit Note" audio type with fields for Site, Issues Found and Follow-up Date can be filled in from a spoken note like "Visited the Riverbank site this morning, two loose handrails on level 3, come back next Tuesday".
- It only fills fields that are empty. Anything a person has already entered is never overwritten.
- Dates are worked out in your timezone, so "next Tuesday" becomes a real date.
- To help it fill reference fields (such as a link to a person or a site), add a Contextual Include: choose a Content Type and optionally narrow it with Filter Keywords, If matches rules and Page Size. Records whose names are mentioned in the transcript are offered to the AI as matches.
- Extraction needs a transcript, so put Transcribe Audio before it in the list.
Always review extracted values. The AI makes its best guess from what was said; it can mishear names or miss details.
Describe Image
The caption becomes the image's description if no one has written one, and the caption's key words and the tags are added to the image's searchable labels. That means searching your media for "workshop" or "outdoor" can find photos nobody tagged by hand. Use Preferred Tags to steer the tags toward your own vocabulary, such as "match day" or "product launch". The AI still adds other relevant tags. If the image is linked to an event, the event's title is used as context for the caption.
Setting up enrichments on a content type
Enrichments are configured on a content type (definition) that is a kind of audio or image. To add them:
- Open the content type in the definition editor. See Creating Definitions if you need a new audio or image type first.
- Open the Enrichment tab. It only appears for audio and image content types.
- Add an enrichment and choose the Enricher: Transcribe Audio, Extract Field Data or Describe Image.
- Fill in the options that appear for that enricher (Language and Diarize, Contextual Includes, or Preferred Tags).
- Add more enrichments if needed and drag them into order. They run in the order listed, each waiting for the one before it to finish.
- Save the content type.
From then on, enrichments run automatically on each new item created with that content type. They don't run again when an item is edited.
Processing existing items or retrying
To run enrichments on items that were uploaded before you set them up, or to retry ones that failed:
- Go to the list of those items.
- Select the items.
- Choose Reprocess media (in the Developer group of the selection actions).
Reprocessing only runs enrichments that haven't already succeeded on an item, so it won't charge you twice for work that's done.
Enrichment status
Enrichment runs in the background after an item is saved, usually within a minute or two (longer for long recordings). Each item records whether its enrichments are pending, complete or failed. If an enrichment fails or is skipped (for example because your organisation is out of credits), the item is saved normally without the AI results, and you can reprocess it later.
Costs
- Describe Image and Extract Field Data use your organisation's credits. They appear on the Credits tab of Billing as "AI file processing". Each run is a small fraction of a credit to a few credits, depending on the size of the image or transcript.
- If your organisation doesn't have enough credits, these enrichments are skipped rather than run. Add credits and reprocess the items.
- Transcribe Audio and AI tagging of uploaded images currently don't use credits.
Permissions and privacy
- Only people who can edit a content type can add or change its enrichments.
- Enrichment results are stored on the item itself, so whoever can see the item can see its transcript, caption and tags.
- Audio and images are sent to Qik's AI providers for processing. Check this suits your organisation's privacy obligations before enabling enrichments on sensitive recordings.