Seedance "may contain real person" error: why it happens and what is allowed
Why Seedance rejects images and videos with real human faces, what the "may contain real person" error means, and the options ByteDance officially allows.
Read as MarkdownThe error means Seedance found a real human face in a reference image or video, and ByteDance does not let its Seedance models take that as input. ByteDance's ModelArk documentation says it plainly: "Seedance 2.5 and Seedance 2.0 series models do not support directly uploading reference images/videos that contain real human faces." This is intended behavior, not a bug, and resending the same file gives the same result. The allowed paths are references without real faces, text-only prompts, and, for a specific real person, an authorization workflow run by ByteDance itself.
Everything below comes from ByteDance's own documentation, terms and apps, checked on September 29, 2026. These rules change, so check the linked pages before you rely on a detail.
What does "the input image may contain real person" mean?
It is one of Seedance's content moderation errors. ByteDance's ModelArk error code list defines two of them, both returned as HTTP 400:
| Error code | Message |
|---|---|
InputImageSensitiveContentDetected.PrivacyInformation | "The request failed because the input image may contain real person." |
InputVideoSensitiveContentDetected.PrivacyInformation | "The request failed because the input video may contain real person." |
The suggested fix on the same page is "The input image may contain a real person. Please replace it and try again." The two codes cover the images and videos you send as input; the list has no real-person code for text or audio.
In ByteDance's own apps the same rule shows up in different words. Dreamina lists Seedance 2.0, 2.0 Fast and 2.0 Mini with the note "Real human faces are not supported." and rejects an upload with "Face detected in your media. Try another one." Jimeng, the Chinese app, shows "识别到你上传的素材中包含人脸信息，请调整素材后再试试" ("a face was detected in the material you uploaded, please adjust it and try again").
Why does Seedance block real human faces?
ByteDance ties the restriction to the unauthorized use of people's likeness. When Seedance 2.0 came to CapCut, its announcement said the model launched "with firm safeguards against violations of our safety policies and the unauthorized use of peoples' likeness or intellectual property", including "the ability to make videos from images or videos that contain real faces" among the restricted capabilities.
The same line runs through its terms. The BytePlus GenAI Acceptable Use Policy prohibits using the service "to depict a person (living or dead)'s voice or likeness without appropriate consent or rights, including unauthorized impersonation, non-consensual intimate imagery, or deceptive impersonation without disclosure". ByteDance Seed's Seedance 2.0 launch post adds: "If you wish to use real human portraits as subject references for video generation, identity verification or prior legal authorization is required."
Are celebrities, politicians and copyrighted characters blocked too?
The face check is not limited to famous people. ByteDance's documentation describes it as covering "real human faces" in general and does not publish a list of protected people or categories. Any photo of a real person can trigger it, whether the person is a celebrity, a politician or a private individual. The acceptable use policy likewise covers the likeness of any person, living or dead, without consent.
Copyright has its own set of errors. The same error code list includes InputImageSensitiveContentDetected.PolicyViolation with the message "The request failed because the input image may be related to copyright restrictions." There are matching codes for input text, video and audio, and for the generated video and audio. CapCut's announcement also lists "advanced technologies designed to block the unauthorized generation of intellectual property" among the Seedance 2.0 safeguards, and the Dreamina Community Guidelines state: "Content that infringes someone else's intellectual property rights is not allowed on our platform".
What does ByteDance officially allow instead?
ByteDance's guide Create portrait videos with Dreamina Seedance models, updated September 28, 2026, lists three options on its ModelArk platform:
| Option | What ByteDance says |
|---|---|
| Trusted model outputs | "The original face-containing outputs generated by certain models under this account can be used as input assets to call Seedance 2.5 and Seedance 2.0 series models again for secondary creation, without triggering input moderation blocking." |
| Preset digital characters | "The platform presets a digital character library, providing creators with free, compliant, and diverse portrait assets." |
| Authorized real-person assets | "Supports using authorized real-person portrait assets for video generation." |
Each option has conditions:
- Trusted outputs cover face-containing videos from Seedance 2.5 and the 2.0 series, their last frames, and images from Seedream 5.0 lite text-to-image, for 30 days after generation. The guide warns: "Only outputs under the same account are trusted; cross-account use is not supported." and "Only original model outputs are trusted; they cannot be used after secondary editing or after the validity period has expired."
- Digital characters are for "scenarios that require real-human-style faces but do not need to specify a particular person". Each character is passed to the model as an
asset://ID. - Authorized real people go through ModelArk's real-human asset library. The person has to consent and authorize your account. The documentation states: "Real-human assets must be used only after the actor's consent and authorization." The requesting account "must complete real-human or enterprise authentication".
Can I use these options on SeedRouter?
Only the options that do not need an asset ID. SeedRouter serves Seedance 2.5 and Seedance 2.0 with its Fast and Mini versions, and its API takes media as public URLs only. The Seedance 2.5 API reference says "Base64, data: URLs, asset:// IDs and multipart uploads are not accepted." ByteDance's digital characters and authorized real-person assets both need asset:// IDs, so a project that must show a specific real person should run that workflow on ByteDance's own platform.
On SeedRouter, the options that work are:
- Text-to-video. Describe the person in the prompt instead of sending a photo. ByteDance's real-person error codes cover input images and videos, not text.
- References without a real face. Illustrated or 3D characters, products, places, and footage where no real face appears.
- Other media types. Seedance can take audio references and videos without people alongside your prompt; see the Seedance API guide for how references are sent.
A task stopped by the face check ends as failed with an error object, and failed tasks are not charged. The Seedance 2.0 API reference lists the error format.
Frequently asked questions
Can I get the real person restriction removed from my account?
Not as a setting. ByteDance's documented way to use a specific real person is the authorized real-person asset library on ModelArk, where that person verifies and gives consent.
Does retrying the same image help?
No. The error reflects what is in the image, so the same file is rejected again. ByteDance's suggestion is to replace the image.
Is it only Seedance 2.0?
No. ByteDance's portrait guide names both Seedance 2.5 and the Seedance 2.0 series, and Dreamina shows the same "Real human faces are not supported." note on Seedance 2.0, 2.0 Fast and 2.0 Mini.
Are AI-generated faces allowed?
ByteDance's documentation does not say how its check treats realistic AI-generated faces. The options it documents are its own trusted outputs, its digital character library and authorized real people.
