Evergreen Tools
|
← Back to AI Tools

TwelveLabs Video Understanding API

Marengo turns video into searchable embeddings while Pegasus generates written answers straight from frames, speech and sound

Tool Interface

Interactive tool will be available soon

Features

  • ✓ The free plan includes 600 minutes (10 hours) of indexing with no credit card required
  • ✓ Marengo handles multimodal retrieval, Pegasus handles video-to-text analysis
  • ✓ Developer is pay-as-you-go: $2.50/hour of video indexing and $4 per 1,000 search queries
  • ✓ The Analyze API bills $1.75 per hour of input video and charges output by token
  • ✓ A single index holds up to 10 hours and 100 videos with 5 concurrent indexing tasks on Free

How to Use

  1. Create a TwelveLabs account — the Free plan grants 600 minutes of indexing immediately
  2. Upload video to build an index and let Marengo generate multimodal embeddings
  3. Query it with the Search API or Jockey in natural language to locate exact moments
  4. Run Pegasus over the selected clips to produce summaries, tags or structured fields

FAQ

What happens when the free allowance runs out?

Free is a cumulative 600-minute indexing allowance, and the docs note the amount does not decrease even if you delete indexes or videos. Once spent you move to the Developer plan (pay-as-you-go with a credit card) or negotiate committed-use Enterprise contracts.

How exactly is the Developer plan billed?

Per the official pricing page: video indexing $2.50/hour, embedding infrastructure $0.09/hour monthly, the Search API $4 per 1,000 queries; the Embed API bills per million tokens at $0.260 for video, $0.065 audio, $0.080 image, $0.200 text and $0.162 document; the Analyze API charges $1.75 per hour of input video.

Do free-tier indexes expire?

Yes. Free-tier indexes are retained for 90 days from creation, with a 10-hour duration limit, 100 videos per index and 5 concurrent indexing tasks. After upgrading to Developer, any index still within that window automatically gains unlimited access.

When should I use Marengo versus Pegasus?

Marengo is the embedding model for retrieval and classification questions like which minute contains this scene. Pegasus is the video-language model for generation questions like what conclusions can be drawn from this footage. Most pipelines use Marengo to locate moments and then hand the clip to Pegasus for the write-up.

How long can a video be?

The pricing page lists Developer limits of a 10,000-hour index holding 100,000 videos with 25 concurrent indexing tasks, versus 10 hours per index on Free; the homepage also highlights handling a single 4-hour video in one API call.