EDIT / UPDATE (later on 2026-07-08): One of the stuck jobs finally completed after ~3h 23m. progress stayed at 0 the entire time, then jumped straight to 100. So this is a severe MULTI-HOUR render-queue backlog, not a permanent failure — jobs do eventually drain. (Normal render time earlier the same day was ~80-90s.) Question still stands: is anyone else seeing multi-hour Sora 2 render times today?
Since ~07:00 UTC on July 8, 2026, all our Sora 2 API video jobs are severely backlogged.
POST /v1/videos is accepted normally (returns a valid id, status “in_progress”)
but progress stays at 0 for hours before it renders (one job took ~3h 23m; earlier the same day the same job took ~80-90s)
while stuck, expires_at stays null and error stays null
Tested on TWO separate OpenAI accounts, and with a minimal request via direct curl:
{“model”:“sora-2”,“prompt”:“a red cube rotating”,“seconds”:“4”,“size”:“720x1280”}
→ same result on both: progress 0 for hours.
The status endpoint (GET /v1/videos/{id}) also intermittently returns HTTP 503
{“error”:{“message”:“Server error”,“type”:“server_error”}}.
Note: the stall started around 16:00 KST (07:00 UTC). status.openai.com shows nothing for Sora (Sora isn’t a tracked component there).
Is anyone else seeing multi-hour Sora 2 render times / progress stuck at 0 today? Is this a known capacity/queue issue as the API winds down toward the Sep 24 sunset?
Based on what you've shared, this looks more like a render queue backlog than a permanent failure or a request formatting issue.
That said, there may not be much you can change on the client side. A few things that are worth trying:
Avoid aggressively resubmitting the same job, since queued jobs may still complete.
Poll GET /v1/videos/{id} with exponential backoff.
Treat intermittent 503 responses as retryable.
Set an internal timeout or alert threshold based on your normal expected render time.
Based on the information available, I don't see enough evidence to conclude this is related to the September 24 Sora API sunset. While the sunset confirms the API is being phased out, this behavior is more consistent with a temporary service issue.
For the whole day, all Sora-2 videos I’ve tried to generate through my app have been stuck at 0% with error=null. We use the OpenAi SDK. Why would this happen for all generations? I can’t even kill the jobs, they return an HTTP 400: “message”: “Video is still being processed”,
“type”: “invalid_request_error”
My longest job it’s been in_queue for almost 12 hours
What am I doing wrong? Is the API failing? Is this related to the upcoming deprecation of Sora?
Hi, that does sound frustrating (as someone powerless over the situation – instead of someone tasked with immediately replicating, escalating, and solving such API issues causing outage for users of the platform services).
I’ll let you know when this job I sent, via the platform site (where there is no argument it should work), transitions from already not meeting timely expectations, to sounding and being frustrating, to the obvious serious platform-wide outage that OpenAI must escalate to infra and platform teams and mark as an outage in the status - instead of shrugging off with disappointing placating inaction.
