AI Analysis

Learn how to use the machine learning tools in the WrangleBot engine.

WrangleBot offers a machine learning interface, that depending on your implementation, will be sending a request to a cloud ai model. If you are using Cloud Sync, you will be able to use "DeepVA" and "Aleph Alpha" on your media assets.

DeepVA vs Aleph Alpha

We offer two models at the moment for visual analysis.

DeepVA is a analytical machine learning model, that can detect 1500 object classes, faces, genders, age among many other attributes. It can also detect landmarks and celebrities.

Aleph Alpha is a generative model, that can react to images, describe or tag them, make assumptions about the image etc. It halucinates more often than DeepVA, but due to it's generative design can accurately describe frames including categorizing moods.

Request Analysis via MetaFile Layer

The easiest way to analyse MetaFiles is to supply the thumbnail UUIDs to the function exposed via the query interface.

const options = {
  engine: "deepva";
  frames: ["<thumbnail_uuid>"]
}

// or

const options = {
  engine: "aleph-alpha",
  frames: ["<thumbnail_uuid>"]
  prompt: "Q: What can I make with these ingredients? A:",
  max_tokens: 16,
  temperature: 0.7,
  stop_sequences: ['\n']
}

await metafile.query.analyse(options);

Request Analysis via Thumbnail Layer

If you prefer you can also query Thumbnails of a MetaFile and analyse the query result of the previous fetch request.

const options = {
  engine: "aleph-alpha",
  prompt: "Q: What can I make with these ingredients? A:",
  max_tokens: 16,
  temperature: 0.7,
  stop_sequences: ['\n']
}

await metafile.query.thumbnails.many({
    $ids: [...]
}).analyse(options)