Seedance 2.0 is ByteDance's AI video model, and Vixxxen runs it across three modes: text-to-video, image-to-video, and video edit. This guide covers what each mode is for, the durations and resolutions actually offered, and how to prompt each one well.
Three modes: text-to-video, image-to-video, and video edit — each suited to a different starting point.
Text-to-video and image-to-video: 720p or 1080p, 5/10/15 second durations. Video edit: 480p, 720p, 1080p, or 4k, 4–15 second durations, and the most strictly moderated mode.
What Seedance 2.0 is
Seedance 2.0 is an AI video generation model built by ByteDance. It is not a Vixxxen model — Vixxxen integrates it as one of the video generation options available in the app, alongside other supported video models. Seedance 2.0 is known for realistic motion, decent prompt adherence, and support for editing existing video rather than only generating from scratch.
Because it's a third-party model, its behavior, availability, and content rules are set by ByteDance and the hosting provider, not by Vixxxen. That matters most for the video edit mode — see the section below.
The three modes and what each is for
Text-to-video
Generates a new clip from a written prompt alone, no starting image or video required. Best when you're creating a scene from scratch and don't already have a still to animate.
Image-to-video
Animates an existing still image, using it as the starting frame. Best when you already have a strong, approved image and want to add motion while preserving the subject and composition.
Video edit
Modifies an existing video — for example, swapping the subject using a reference image, or matching motion between two sources. Most powerful mode, but also the most heavily screened. See the content filter guide before relying on it.
Durations and resolutions on Vixxxen
These are the actual options exposed in the Seedance 2.0 integration, not general model capabilities:
Mode
Resolutions
Durations
Text-to-video / image-to-video
720p, 1080p
5, 10, or 15 seconds
Video edit
480p, 720p, 1080p, 4k
4–15 seconds
Higher resolutions and longer durations cost more credits. If you're testing a prompt, generate a short clip at 720p first, then re-run the winning prompt at your target resolution and length.
Practical prompting tips by mode
Text-to-video
Describe subject, action, setting, and camera framing in that order — the model weights early details more heavily.
Keep the action simple. One clear motion (walking, turning, a hand gesture) renders more reliably than a sequence of actions.
Specify camera behavior explicitly if you want it: "slow push in," "static shot," "handheld." Leaving it out gives the model more freedom than you probably want.
Image-to-video
Start from your best still, not a rough draft. Whatever imperfections exist in the source image tend to persist or amplify in motion.
Prompt for motion only, not scene changes. The image already defines subject, outfit, and background — re-describing them can conflict with the source and cause drift.
Short, physically plausible motion (a turn, a step, hair or fabric movement) holds up better than large or fast movements.
Video edit
Scope each edit to one change: swap the subject or the background, not both in the same request.
Say "match the motion" rather than describing new movement — it reliably preserves the source video's motion.
If you need multiple changes, run them as sequential edits instead of stacking them into one prompt.
Video edit is the strictest-moderated mode on Seedance 2.0. If a job gets stuck around 90% progress and then fails, that's a content flag, not a bug — read the Seedance 2.0 content filter guide for what triggers it and how to avoid it.
Why video edit needs its own guide
ByteDance and the hosting provider run stricter automated screening on video edit jobs than on text-to-video or image-to-video, mainly around realistic face swaps and identity replacement. Vixxxen doesn't control or add to that screening. Flagged jobs are auto-refunded, but they can sit at ~90% progress for up to 20 minutes before failing, which looks like a stall rather than a rejection.
If your video edit jobs keep failing after a long wait, don't assume it's a platform bug. Read the Seedance 2.0 content filter guide for real pass/fail prompt examples and how to scope an edit so it passes.
Frequently asked questions
What is Seedance 2.0?
Seedance 2.0 is ByteDance's AI video generation model. On Vixxxen it powers text-to-video, image-to-video, and video edit workflows for creator content.
What resolutions and durations does Seedance 2.0 support on Vixxxen?
Text-to-video and image-to-video jobs support 720p or 1080p at 5, 10, or 15 second durations. Video edit jobs support 480p, 720p, 1080p, or 4k at durations from 4 to 15 seconds.
Which Seedance 2.0 mode should I use?
Use text-to-video to generate a new clip from a written prompt, image-to-video to animate an existing still, and video edit to modify an existing video, for example swapping the subject or matching motion to a reference image.
Why does Seedance 2.0 video edit fail more often than the other modes?
Video edit is screened more strictly than text-to-video or image-to-video because it can involve identity replacement. See the Seedance 2.0 content filter guide for what triggers a flag and how to prompt around it.
