The scene looks still or only zooms slowly
Name a visible action with a start and finish: a skater approaches, leans into a turn and stops. State where the subject travels in the frame, then choose a camera that keeps the movement visible.
Words such as cinematic, beautiful and dynamic describe a look without specifying an action. Remove slow motion or gentle push-in if you want normal-speed movement. Keep enough space in a starting image for the action to happen.
First watch the complete video, not just its poster or a short gallery preview. Note the moment when the subject should start moving and what actually changes. If the action is vague, change only the action sentence for a planned comparison: replace “a dynamic skater” with “the skater moves from the left edge to the bench on the right and stops beside it.” Keep the camera clause, reference image and output settings unchanged. Check displacement relative to the scene, not only whether the whole image gets larger. The motion template below is a suggested starting point, not a tested fix.
One continuous five-second shot at normal speed. A small clay creature crouches, springs from the left platform to the right platform, and lands with a visible squash. Fixed wide camera showing both platforms and the entire jump. Keep the creature’s color and face consistent.The camera zooms or moves when you wanted a fixed shot
Separate two observations: did the subject perform the action, and did the framing move? Pick two background landmarks and compare their positions at the start, middle and end, then watch the whole clip. Record whether the view tightens, drifts sideways or changes angle. A subject getting larger in the frame does not by itself show that the intended action happened.
Read only the camera instructions in your prompt. A push-in, orbit, tracking shot or handheld description can conflict with a request for a fixed view. For one planned test, replace the camera clause with the clause below; leave the subject action, image and settings unchanged. This is an untested diagnostic edit. Fixed-camera wording and continuous face tracking are not guaranteed controls, and this workspace has no dedicated camera-path or face-lock setting.
Fixed wide camera. Keep the framing unchanged from start to finish, with the entire action area visible. No camera pan, tilt, zoom, dolly or orbit.The image is present, but the opening frame is wrong
Check the selected mode and upload preview before rewriting the prompt. Image to video uses one image as the first frame. Reference to video uses one image to guide the scene without requiring it to be the opening frame. Text to video does not attach an image. A mode mismatch is a setup issue to correct before testing wording.
If the goal is to animate the supplied composition, select Image to video and confirm that the intended image appears in the upload area. Reuse settings may ask you to add the image again. If the goal is a new shot guided by the same subject, choose Reference to video instead. Neither mode locks a character perfectly or accepts a first-and-last-frame pair here.
Treat a mode correction as its own change. Keep the image file and action wording fixed, review Match image or the selectable ratio, and note any settings that change automatically. Check the new quote. Because changing modes can also change the aspect-ratio setting, do not describe that result as a controlled prompt-only comparison.
| What you need | Select | Confirm before submission |
|---|---|---|
| The video begins from the uploaded composition | Image to video | One starting image is visible; Seedance 2.5 uses Match image, while other models use the selected supported ratio. |
| The image guides a new scene or shot | Reference to video | One reference image is visible; review the chosen output ratio. |
| A scene described entirely in words | Text to video | The prompt describes the scene; no image is required. |
The subject, hands or product changes shape
Reduce the number of actions, subjects or camera changes. Choose a clear source image with an unobstructed subject and avoid asking it to transform while also preserving its exact shape.
For products, inspect labels, edges and reflections through the whole clip. Add exact slogans and prices in an editor; a successful generation does not guarantee accurate lettering.
Choose one suspected conflict for the next planned revision. For example, remove a transformation request while keeping the source image, camera and settings fixed. Compare the same detail at the start, middle and end and during the fastest movement. If you instead crop or replace the image, record that as a separate input change; do not change the image and several prompt clauses together and assume you know which edit helped.
Prepare a product imageKeep character references consistentThe image will not upload, or the frame is wrong
SuperSeedance currently accepts one still JPG, PNG or WebP up to 10 MB. Both dimensions must be 300–6,000 pixels, with width divided by height between 0.4 and 2.5. These are this studio’s upload limits; other Seedance interfaces may accept different formats or inputs.
If your photo is HEIC or HEIF, export a JPEG or PNG copy; changing the filename extension does not convert the image. On a Mac, open it in Photos or Preview, choose File → Export and select JPEG or PNG. Check the exported file’s size and dimensions before uploading it.
Seedance 2.5 image mode uses Match image for its aspect ratio. Prepare the starting image in the composition you want. Text and reference modes offer selectable ratios; check the chosen ratio before submitting.
Writing 9:16 in the prompt does not replace the ratio setting or a prepared vertical first frame. For image mode, crop the source before uploading; for text or reference mode, choose 9:16 in the settings. Inspect the downloaded file’s dimensions and the subject throughout the clip.
An upload rejection does not require another generation. Check the file type, byte size, dimensions and ratio first, then inspect the accepted upload preview. If you crop the source, save it as a new version so you can distinguish a composition change from a prompt change.
Follow the image-to-video stepsPrepare a vertical image and 9:16 videoThe video is silent or contains unwanted music
First check player mute, volume and the selected audio output, then play the downloaded original in another player. Review the original task’s Native audio setting: it defaults to off in SuperSeedance. Turning it on now cannot add a track to a video that was already generated without one.
For sound effects without music, keep native audio on and describe the intended sounds. A request for no music is not a guaranteed filter. If the existing visual result works, you can mute or replace its audio in an external editor instead of regenerating the video; muting also removes dialogue and effects.
Set up native audio and diagnose sound problemsWhat the existing completed samples actually show
These are observations from three real completed clips in the eight-example collection, not failed generation jobs. All eight source requests used Seedance 2.5 text to video, six seconds, 720p, 16:9, no audio and Draft off. The recorded review sampled 12 frames per clip; it does not prove every intermediate frame is free of defects.
Open the full clip when comparing it with the prompt. Some gallery previews shorten preparation or settling. None of the revisions suggested in this guide was tested as a before-and-after fix for these clips, and these text-to-video results do not validate an image-to-video workflow.
| Completed sample | Recorded observation | Useful diagnostic question |
|---|---|---|
| Solo Dance | The first half spends longer preparing; the later turn, opened silk and finish are more pronounced. | Am I judging a still opening, a short preview or the complete action? |
| Clay Animation | Landing on the drum, compression, rebound and landing on the cushion are visible. Compression is smaller than requested and there is a pause in the middle. | Is the action absent, or is its timing or amplitude different from the brief? |
| Liquid Metal | The tight coil opens into a large ring; the later motion is mainly rotation. The requested open spiral is not fully achieved. | Did the output perform the requested shape change, or a different kind of movement? |
The task is queued, saving or needs review
Open My Creations to see the saved task and its latest message. You can leave the generation page after submission. Avoid starting a duplicate while the original task’s result is uncertain.
Write down the submission time, current status and exact visible message. A slow response or dropped connection does not confirm failure. If the status is Needs review, keep the existing record and contact support rather than starting a replacement request. Confirmed failed-task credit returns are different from cash purchase refunds; this guide does not promise a return within a fixed number of minutes.
| State | What to do | Credits |
|---|---|---|
| Waiting in queue | Wait, or cancel while the Cancel task button is available. | Reserved; returned if cancellation succeeds. |
| Sending / Generating / Saving | Keep the existing task; saving means the file is being stored. | Reserved; the task is not confirmed failed. |
| Needs review | Contact support with the task ID and visible message. | Charge or return is confirmed after review. |
| Generation failed / Cancelled | Check the message and credit history before trying again. | Reserved credits are returned. |
| Completed | Play or download the result and review the full clip. | The completed generation uses credits. |
Generation is unavailable, or the video will not play
Before submitting, verify your email, select an available model and parameter combination, and check that you have enough credits. At most three active tasks are allowed; a task awaiting review also occupies a slot.
If a completed video will not play in the browser, try downloading it and opening it in a compatible player before generating again. If the file is unavailable, send support the task ID and error message.
Seedance 2.5 generates 1080p files with 10-bit H.265/HEVC encoding, which some browsers cannot play. Try the downloaded original in VLC, mpv or QuickTime Player. If that works and you need a more compatible copy, export H.264 in an external editor; in QuickTime Player on a Mac, use File → Export As and choose Greater Compatibility (H.264). Keep the original: exporting creates a converted file, not a new Seedance generation.
A playback problem alone is not a reason to pay for another generation. If you need a new request, check the balance and current quote first; a retry does not provide free generation.
Check your accountReview the credit estimateKeep a one-change diagnostic record
Copy this worksheet into your own notes before another attempt. Define one visible acceptance check, such as “the subject reaches the right platform” or “both background landmarks stay in place.” Record one change and keep the other inputs fixed where possible. Generations can vary even with the same prompt, so a single improved result does not prove a universal fix.
For a service problem, share the task ID if available, submission time, status and visible error with support. Keep the original task record until the outcome and credit entry are clear. For a creative revision, compare the complete clips against the same acceptance check before deciding whether to spend more.
Task ID / submission time / timezone:
Current status and visible message:
Mode / model / image filename (if used):
Prompt version; subject action and camera instruction:
Duration / resolution / aspect ratio / audio / Draft:
Quoted credits and credit-history entry:
Expected visible result:
Observed result and video time(s):
One planned change; other inputs and settings held fixed:
Maximum additional attempts / credit budget:
After review: met / partly met / not met?
Next step: keep / revise within budget / stop / contact supportWhen to stop and avoid another paid request
Set your own attempt and credit limit before testing. Review each completed result before authorizing the next. Stop when that limit is reached, when the same acceptance check still fails after your planned comparison, or when you cannot name a specific next change. Return to the shot plan or edit the usable footage instead of repeating an unchanged request indefinitely.
| Current condition | Next action |
|---|---|
| Original task is Sending, Generating, Saving or Needs review | Check the existing task. Do not create a duplicate to resolve uncertainty. |
| Image rejected, wrong input mode or required image missing | Correct the input and inspect its preview before reviewing a new quote. |
| Account, available settings or sufficient balance cannot be confirmed | Resolve the displayed requirement first. Watching examples and editing a prompt do not submit a task. |
| Task completed but the clip will not play | Try the download and a compatible player before deciding the generation itself is unusable. |
| Planned attempts or credits exhausted, or no clear change to test | Stop generation, retain the records and revise the brief offline. |
Questions and answers
Do I get credits back if a video looks wrong?+
A completed generation uses credits even if its movement or appearance is not what you wanted. Reserved credits are returned when a task is confirmed failed or successfully cancelled.
Can I cancel a running generation?+
Only a task still waiting in the local queue can be cancelled. Once it is being sent to the model or is running, cancellation is not available here.
Should I resubmit when the connection drops?+
Check My Creations first. If the original request needs review, do not assume it failed or create another paid request while its result is being confirmed.
Is Try again free?+
No. It starts a new generation with its own credit requirement. Fix the reported issue and review the settings before trying again.
Is a 480p video automatically a draft?+
No. Resolution and draft mode are separate settings. Draft is available only at 480p; changing resolution for a new task does not upgrade an existing video.
Why is the subject not moving even though the camera zooms?+
The camera and subject have separate instructions. Check whether the subject changes position or pose relative to the scene. For a planned comparison, change only an unclear action sentence into one visible action with a direction and endpoint, while keeping the camera and other settings fixed. This is a diagnostic suggestion, not a guaranteed fix.
Will a fixed-camera prompt prevent every unwanted zoom?+
No guarantee is established here. Remove contradictory camera directions and test one camera clause at a time within your budget. Compare background landmarks and framing throughout the result. The workspace has no dedicated camera-lock or face-lock control.
Why does my reference image not become the first frame?+
Check the selected mode. Image to video uses one starting image; Reference to video uses one visual reference without requiring it to be the opening frame. Confirm the upload preview, and recheck the aspect ratio and credit quote when switching modes.
Are the clay and liquid-metal examples failed jobs?+
No. Both are completed text-to-video samples. Their recorded motion differs from parts of the requested prompt. A completed creative mismatch and a confirmed task failure have different credit outcomes, and no tested repair is claimed for those examples.
How many times should I regenerate before stopping?+
There is no tested number that guarantees a good result. Set an attempt and credit limit first. Pause if the original task is unresolved, if a required input or balance is missing, if your planned comparison still fails the acceptance check, or if you have no specific next change. Each new completed generation consumes credits.
Sources
The sample observations below come from completed Seedance 2.5 text-to-video clips generated on 21 October 2026. Diagnostic prompt edits are suggestions to test, not verified fixes, camera-lock guarantees or measured success rates.