← Back to blog

Seedance "Flagged as Sensitive" (E005): Why Photos With Faces Fail

What Seedance's E005 error means, why a reference photo with a close-up face triggers it, the framing that passes, and why retrying the same image does not help.

Seedance's E005 error reads: "The input or output was flagged as sensitive. Please try again with different inputs." It comes from ByteDance's own filter, so it follows the model to every platform that runs it. The trigger we have confirmed is not an explicit prompt. It is a face shown in close-up in a reference image. The same person framed from the waist up goes through.

TL;DR

What the error says, and the one that looks like it

Two different Seedance failures get confused with each other.

Message What it means
"The input or output was flagged as sensitive. Please try again with different inputs. (E005)" The content filter rejected an input (most often a face in a reference image) or the result
"The request failed because the output video may be related to copyright restrictions." A separate check on whether the result resembles protected material

The fixes are different, so read the whole message before changing anything.

What triggers E005: faces in close-up

We ran the same person through Seedance 2.0 and 2.5 at 480p with different reference framings:

Reference image Result
Close-up portrait, photo Rejected, every run
Close-up portrait, stylised 3D render Rejected
Waist-up, face about a quarter of the frame height Passed, identity kept
Full-length figure, small face Passed, identity kept
No reference image (text to video) Passed

Two things surprised us. Stylisation did not help: a stylised 3D render of a face in close-up was rejected just like a photo. And the filter reacted to the reference, not to the result: outputs that moved in to a close-up of the same face came back fine.

Our production logs point the same way. Between June and September 2026, Seedance runs on 8frame that carried a reference image failed more than twice as often as runs without one.

How to fix it

  1. Reframe the reference. Use a waist-up or wider shot. Aim for the face at roughly a quarter of the frame height or smaller.
  2. Put the close-up in the prompt, not in the reference. Keep the reference wide and ask for the push-in or the close shot in the text.
  3. Do not retry the identical image. It fails the same way. Change the image, not the attempt count.
  4. Reuse one good reference across shots. If a waist-up reference passes, use that same image for every shot of that person so the identity stays consistent.
  5. Check the prompt last. If the reference is already wide and the run still fails, then look at the prompt: violence, nudity and named real people are what these filters exist to catch.

If you need a face-forward talking shot

A tight talking-head shot is exactly the framing Seedance's filter dislikes, so it is the wrong model for that job. Tools built for a portrait plus an audio track, such as OmniHuman, start from a face by design. The trade-offs are covered in best AI talking avatar generators.

The copyright variant

"The output video may be related to copyright restrictions" is a different check. Rewording helps only if you remove whatever resembles a protected work: named characters, franchise titles, logos, or a reference frame taken from a film. It is not related to faces, and reframing a reference will not fix it.

What it costs you on 8frame

Nothing in credits. When a generation fails, 8frame returns the credits it cost in the same step that marks it failed. The cost is time: a Seedance run takes minutes (see how long AI video generation takes), which is the real reason to fix the reference rather than retry. For what a successful run costs, see Seedance 2.5 pricing.

FAQ

What does Seedance error E005 mean? ByteDance's content filter flagged an input or the output as sensitive. With reference images, the trigger we have confirmed is a face in close-up.

Why does Seedance reject my photo but accept a text prompt? The filter reacts strongly to faces in reference images. In our tests, the same person passed when shown waist-up or full-length, and text-to-video without a face reference worked on both 2.0 and 2.5.

Does retrying fix E005? Not with the same image. Since August 2026 the rejection has been consistent. Change the framing of the reference instead.

Is this only a Seedance 2.5 problem? No. Close-up references failed on Seedance 2.0 and 2.5 alike in our tests.

Do I lose credits when Seedance fails on 8frame? No. Failed generations are refunded automatically.


Frame the reference wide, ask for the close-up in the prompt. The 8frame canvas is free and unlimited, and generation is paid from $19/month.

Related articles

guideAI Generation Refusal Messages, Decoded: What Each Error Actually MeansguideUpload Limits for AI Video and Image Models in 2026: Size, Resolution, Formatprompt guideAI Prompt Length Limits by Model in 2026: Kling, Seedance, Wan, Hailuo, Midjourney

Make it
move.

Stay in the loop

Be the first to hear about our launch and get product updates