How to generate AI video in n8n, Make and Zapier

Video takes longer than one HTTP call should wait for, so the KreatorFlow API works in jobs. You start a job with one request, and you either check on it or get told when it is done. Every automation tool can do both: an HTTP step to start and check, and a webhook trigger to be called back.

The models are ByteDance's Seedance for video and Seedream for images, run by KreatorFlow, an official BytePlus partner. Every create response includes the quoted token cost, so a workflow can log or cap spend per run.

Use the HTTP Request node. n8n's docs describe Bearer auth as a generic credential: create a Bearer Auth credential, paste your KreatorFlow key, and n8n sends Authorization: Bearer <key> for you. Set the method to POST, the URL to https://kreatorflow.ai/v1/videos and the body to JSON.

To wait, the neatest option is the Wait node set to resume On Webhook Call. n8n's docs say it creates a resume URL for each execution, available as $execution.resumeUrl. Put that expression in webhook_url on the create request, then add a Wait node after it: the workflow pauses until KreatorFlow posts the finished job to that URL. Set the Wait node to accept POST and give it a time limit as a safety net.

If you prefer polling, add a Wait node set to a short time interval, then an HTTP Request node that GETs https://kreatorflow.ai/v1/jobs/{{ $json.data.job.id }}, and an IF node that loops back until the status is completed, failed, canceled or blocked.

Finish with an HTTP Request node that GETs /v1/jobs/{job_id}/output with the same credential and the response format set to file, then send the file wherever it needs to go.

Use the HTTP app's Make a request module. Make's documentation lists its settings: authentication, URL, method, headers, query parameters and body content type. Set the method to POST, the URL to https://kreatorflow.ai/v1/videos, add a header Authorization with the value Bearer followed by your key (or use the API key authentication option), choose JSON as the body type and map your prompt into the body.

To wait, start a second scenario with a Webhooks custom webhook trigger and put its URL in webhook_url on the create request. When KreatorFlow calls it, that scenario gets the finished job and can download the output with another Make a request module (GET /v1/jobs/{job_id}/output, same Authorization header).

For a single scenario, use a Sleep module and a repeating Make a request that GETs /v1/jobs/{job_id} until the status is final. Mind Make's module timeout (Make documents a setting from 1 to 300 seconds): keep each request short and let the loop do the waiting.

Use Webhooks by Zapier. Its Custom Request action sends raw JSON exactly as you type it, and custom headers go in the Headers section, so it suits an API call with a Bearer header. Choose method POST, URL https://kreatorflow.ai/v1/videos, header Authorization: Bearer <your key> and Content-Type: application/json, and write the JSON body with your prompt mapped in.

To get the result, build a second Zap that starts with Webhooks by Zapier's Catch Hook trigger. Zapier documents that Catch Hook accepts POST requests and parses the body, so the finished job arrives as fields. Put the Catch Hook URL in webhook_url on the create request, then add a Custom Request step in the second Zap that GETs /v1/jobs/{job_id}/output with the same header.

Zapier's Custom Request does not validate your JSON, so test the body once with a fixed prompt before mapping fields into it.

A POST with a JSON event: type is job.completed or job.failed (job.<status> in general), and data.job holds the same job object as GET /v1/jobs/{job_id}, including id, status, errorCode and, for a finished job, the output link. The webhook URL must use HTTPS.

Each call carries X-KreatorFlow-Event-Id, X-KreatorFlow-Event-Type, X-KreatorFlow-Job-Id, X-KreatorFlow-Timestamp and X-KreatorFlow-Signature headers. Automation tools rarely verify signatures, so treat the webhook as a notice: read the job id from it, then fetch the job and the output with your own key, which only works for your own jobs.

Anything that starts with a row, a message or a schedule. A few workflows that suit the API:

Every create response includes quoted_token_cost, so log it or check it before you continue. Send an Idempotency-Key header built from the row or record id, so a tool that retries a failed step never makes and bills the same clip twice. Prepaid tokens are also a hard cap: when the balance runs out, the API answers insufficient_balance instead of charging more.

Test the workflow with short clips at a lower resolution, then raise the settings once the output looks right. The quote in each create response shows what a setting costs before it runs.

If you want an assistant to decide what to make, use KreatorFlow's MCP server instead of building the HTTP steps yourself. Claude, ChatGPT and other MCP clients get tools such as generate_video and get_job and call them on your account.