image_not_fetchable error: what it means and how to fix it

image_not_fetchable means Sume could not fetch or mirror the input media safely. The documented fix is to check that the input is a public HTTPS image URL, then retry, or contact support with the request id if it keeps happening.

The rows come from Errors and credits, read 2026-09-29. The same table lists input_media_unreachable and storage configuration errors under the same guidance.

What counts as a fetchable URL?

These routes take a URL, not file bytes. Host the image somewhere the open internet can reach, then pass its URL.

A URL that works on your laptop can still fail here. Sume fetches the file from its own servers, so anything that depends on your network, your login, or a browser session is invisible to it. A link that redirects to a sign-in page is the same problem, so use a URL that serves the image without one.

What should I check first?

Open the URL in a private browser window, where you are signed out. Then check these:

• The link is https://, not http://.
• The host is not localhost or a private-network address.
• It opens without a login, cookie, or VPN, so it is not a private share link.
• It points at the image itself and not at a web page that shows the image.
• It has not expired, if it came from a signed URL such as an earlier Sume /v1/images result.

What should I send to support?

The docs ask for the request id. Log the request id from every failed response, together with the route and the URL you sent, so the report is one message instead of a thread.

Keep the image URL out of any log that is shared publicly if it is a signed link, since the signature is what grants access. Failed jobs also expose public error metadata such as category, stage, and retryability, so read those fields on the job before deciding to retry.

Does retrying help?

For this code the guidance is to check the URL first and then retry. A retry of a URL that is still private will fail the same way, so fix the input before you resend. Do not assume a retry fixes it. For the queue-related codes on the same page, the guidance is different, so read the table row for the exact code you got.