Make videos searchable with description, tags, and speech

View Markdown

Index video content with descriptions, tags, and speech-to-text to enable fast searching and discovery.

This workflow generates a description (with tags), transcribes speech, and uses prompt to generate a domain-specific set of keywords.

Then we can save this intelligence data to our database, and use it to populate our search indexes.

Workflow

[
  {
    "kind": "description"
  },
  {
    "kind": "speech"
  },
  {
    "kind": "prompt",
    "prompt": "Return a list of any locations visible or mentioned in this video. Use structured JSON: {"locations": [<string>]}. Prefer city and region names over specific addresses."
  }
]

Create Task

const task = await ittybit.tasks.create({
  url: "https://ittyb.it/sample.mp4",
  kind: "workflow",
  workflow: [
    {
      kind: "description",
    },
    {
      kind: "speech",
    },
    {
      kind: "prompt",
      prompt: "Return a list of any locations visible or mentioned in this video. Use structured JSON: {"locations": [<string>]}. Prefer city and region names over specific addresses.",
    }
  ]
})

On this page