n8n HTTP Request timeout: what to do when a job takes minutes

The n8n HTTP Request node's Timeout option sets how many milliseconds the node waits for the server to start its response, and aborts the request when that time runs out. Raising it helps a slow endpoint that does answer; it is the wrong fix for an API job that runs for minutes. Submit the job so the API answers at once with a job id, then wait in n8n with a Wait node and poll the job's status.

The n8n facts come from n8n's HTTP Request node and Wait node docs. The Sume facts come from Jobs and results and Video generation. All were read on 2026-09-29. Sume has no official n8n node; the workflow below is plain HTTPS calls.

Where is the Timeout setting in the HTTP Request node?

In the node's options, select Add Option and pick Timeout. n8n describes it as how long the node waits "for the server to send response headers (and start the response body)", entered in milliseconds. Requests that exceed it are aborted for the initial response.

The Response option matters when a call fails. By default the node returns success only on a 2xx response; Never Error returns success whatever the code. Include Response Headers and Status returns the headers and status code as well as the body, which you need if you want to read a rate-limit header.

Why doesn't a longer timeout fix a slow video API?

Because the job, not the HTTP call, is what takes minutes. Sume's video docs say generation is asynchronous and typically takes 30 seconds to several minutes, depending on the model and parameters. POST /v1/videos returns a job id and a polling URL immediately; you then poll GET /v1/videos/{jobId} until the status is completed.

Where a Sume job submit takes a blocking sync mode, wait_timeout_seconds is clamped to 0–30. It bounds the HTTP wait, not the job. When it runs out, the response is still a 2xx carrying the job id, and the docs say you must keep polling and must not submit a new paid job for the same intent. So no n8n Timeout value turns a minutes-long job into one request.

How do I call a slow API from n8n without timing out?

Split the work into a submit and a wait loop. The Wait node pauses the execution and, per n8n, offloads the execution data to the database until the resume condition is met, so nothing holds an HTTP connection open while the job runs.

• HTTP Request (POST) to https://api.sume.com/v1/videos with Authorization: Bearer from an n8n credential and an Idempotency-Key built from your item, not a random value.
• Keep the returned id. A 2xx here means the job exists, not that it finished.
• Wait node, Resume: After Time Interval. Sume's docs suggest a reasonable polling interval such as 30 seconds.
• HTTP Request (GET) to https://api.sume.com/v1/videos/{id}, then an If node on status.
• completed: download unsigned_urls[0] (the job's /content URL) with the same Bearer header. failed: read the error field. cancelled: stop. Anything else: loop back to the Wait node.

POST https://api.sume.com/v1/videos
Authorization: Bearer $SUME_API_KEY
Content-Type: application/json
Idempotency-Key: order-123-video-v1

{ "model": "sume/auto", "prompt": "A vertical product clip on a desk, natural light", "aspect_ratio": "9:16", "duration": 5 }

Which waits apply to each step?

Each wait below bounds a different thing. Only the first one is set in n8n.

What if the node already timed out on a create call?

The job may exist even though n8n saw an error. Retry the same request with the same Idempotency-Key and the same body: the video docs say a replay returns the original job, so you get its id without a second job. Retrying without a key, or with a fresh one each run, can start a second paid job, and Sume's error docs say not to retry unsafe submits without an Idempotency-Key.

If you would rather be told than poll, pass callback_url and resume the workflow on the webhook instead; n8n AI video workflow: resume a Wait node on a Sume webhook shows that setup. Sume's own caps on each wait are in AI video API timeouts.