Lambda 90-minute timeout: does it change AI video jobs?

Not much: AWS's 90-minute limit covers only asynchronous and event-source-mapping invocations on Lambda Managed Instances, and a Sume job is still best handled as submit, then poll or webhook, rather than one function held open. Sume's own sync wait stays capped at 30 seconds whatever your function's timeout is.

Sume facts are from the Jobs and results and Webhooks docs; the AWS fact is from its announcement, read 2026-09-30.

What exactly did AWS change?

The announcement says Lambda “now supports a 90-minute function timeout for asynchronous and event source mapping (ESM) invocations on Lambda Managed Instances (LMI), a 6x increase from the previous 15-minute limit.” Synchronous invocations keep the existing 15-minute maximum.

How should a Lambda function call Sume?

Submit with mode: "async" and an Idempotency-Key, keep the job id, and return. The submit answers 202 with status_url and result_url at once. Then either take a signed job.completed webhook or poll status_url, honoring next_poll_after_seconds when it is present.

curl -X POST https://api.sume.com/v1/image-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: lambda-demo-001" \
  -d '{"prompt":"Product hero shot of a matte black bottle","mode":"async"}'

curl https://api.sume.com/v1/jobs/job_123/status \
  -H "Authorization: Bearer $SUME_API_KEY"

Can one long-lived function just wait for the job?

It can, but the wait is yours to own; the docs suggest a client-side deadline of about 20 minutes for video. The docs say a client-side timeout does not cancel the job: it keeps running and still bills. If the invocation is retried, reuse the same Idempotency-Key so the retry returns the original job instead of billing a second one.

Does this change the 30-second sync limit?

No. sync and subscribe are the same bounded wait, capped at 30 seconds, and the docs call them the wrong tool for anything that can outlast that. A longer function timeout does not lengthen it; when the budget runs out, keep polling the job id and do not resubmit.