Replicate#

<!-- TODO: expand -->

Replicate is a hosting platform that runs open-source AI models behind a unified API. JSON2Video integrates with Replicate to generate images and short video clips inside a movie render.

When to use Replicate#

• You want to use an open-source image model (SDXL, Flux variants, etc.) that isn't available as a first-party JSON2Video model.
• You're testing a model that's hosted on Replicate before deciding whether to use it in production.
• You want to control the exact model version (Replicate URLs are version-pinned).

How it appears in JSON2Video#

Replicate-backed models are listed in the AI models catalog under image and video categories. Reference them by model name on image or video elements:

{
  "type": "image",
  "prompt": "Aerial shot of a coastal cliff at sunset, cinematic",
  "model": "flux-schnell",
  "duration": 4
}

The available Replicate-backed models change as new ones are added — always check the live AI models catalog for current options.

Bring your own Replicate API key#

To use your own Replicate account (and pay Replicate directly):

1. Get an API token from replicate.com/account/api-tokens.
2. Create a Connection at json2video.com/dashboard/connections with type Replicate and paste the token.
3. Reference the Connection by ID in your element:

{
  "type": "image",
  "prompt": "...",
  "model": "flux-schnell",
  "connection": "my-replicate"
}

When connection is set, JSON2Video forwards the call to Replicate with your token. Replicate bills you for the generation; JSON2Video charges an orchestration fee.

Cost#

See the credit consumption table for JSON2Video-billed usage and replicate.com/pricing for per-model rates when using your own connection.

See also#

• AI models catalog
• Image element reference
• Dashboard — Connections
• Tutorial 9 — AI generated images and videos