Pick the workflow that matches your scene
The AI video generator is a workspace for your own motion idea. A text-led task starts from the scene you describe. A reference-led task uses the input types supported by the selected model, such as an image or video. Check the active controls rather than assuming every model accepts every kind of reference. The available mode determines what the provider receives and which settings affect the result.
This workspace is separate from the named Burning Bridges party effect. The homepage is intended for that focused cup-in-hand performance once its reference and behavior are accepted. This page is for an independently described scene, camera movement or animation. If you need a new still image first, use the image workspace and inspect that result before bringing it into a video task. A still image task does not automatically submit a video.
Describe one clear sequence of motion
A practical video prompt explains who or what is moving, where the scene takes place, what happens first and how it ends. Keep a short clip focused on a manageable action. For example: “An adult in a plain shirt stands in a softly lit lounge, raises one hand in a small greeting, then returns to a relaxed pose; fixed camera, natural movement.” This is a drafting example rather than a demonstrated output or a guarantee of exact choreography.
Separate subject movement from camera movement. “The person walks toward the camera” and “the camera moves toward the person” can produce different framing. Mention which is intended. If a prop must remain visible, describe its role and the hand holding it. Avoid stacking unrelated actions, fast cuts and several scene changes into one short request. A precise sequence is easier to review than a list of competing cinematic adjectives.
Choose references with visible information
A reference photo should show the details you expect the output to preserve. For a face-led animation, use a clear adult face with consistent lighting. For a full-body action, include the relevant limbs, clothing and surrounding space. If a hand or cup is hidden or cut off, the model may invent its position and appearance. A reference can guide generation without guaranteeing an exact reconstruction.
For a supported video reference, review its full duration before uploading it. Check framing, cuts, motion, visible people and sound. A reference that changes subjects halfway through can create an ambiguous request. Only upload material you have the rights to process. Do not use someone else’s likeness to imply actions or endorsements without permission. Selecting a file in the active uploader sends it before Generate, so remove unrelated sensitive information first.
Read duration, resolution and sound settings
The controls shown for a model define the options available for that task. Duration describes the requested clip length; the actual media container can differ slightly because of frame and audio timing. Resolution and aspect ratio affect detail and composition, and a taller frame can show different body coverage from a landscape frame. Decide how you intend to view or share the clip before choosing these options.
Audio behavior depends on the selected model and mode. Do not assume a reference soundtrack will be preserved exactly or that a sound switch exists for every route. If matching music, speech or a particular beat is essential, inspect the finished file with sound. A successful provider status cannot establish that the soundtrack is the expected recording, that its pitch is unchanged or that a gesture lands on the intended beat.
Understand the credit requirement
Review the task cost after selecting the model, input and output settings. Some video workflows charge for reference-video duration as well as generated duration. Additional processing can have its own cost. The site must display the applicable credit requirement before submission; an estimate based only on the output length may omit part of the workflow.
See Pricing for monthly plan prices and credits. Unused subscription credits expire at the end of each billing period and do not roll over. Cancel before renewal to stop future charges. A stored balance does not make a task costless. If you have gifted credits, the balance and the selected workflow still need to be checked. Do not create a second request simply because the first is taking longer than expected, since separate tasks can use separate credits.
Follow the task through to a saved result
Queued and processing states mean the clip is still being prepared. Keep the task reference and use history to check progress after a browser interruption. A failed page request does not necessarily cancel a provider task already submitted. If a result is still pending, avoid interpreting an empty preview as evidence that no generation occurred.
When a task completes, open the entire video rather than checking only its first frame. Examine the face and clothing throughout, then pay attention to hands, held objects and body proportions. Look for objects that change shape, sudden cuts, duplicated limbs or movement that stops abruptly. Watch the ending as carefully as the opening. Listen once without distraction and compare the timing of sound and visible action.
Evaluate timing and make a focused revision
For a gesture-led clip, identify the point where each important action should happen. Compare it with the actual result. A broad resemblance to the reference does not establish precise choreography or beat alignment. Note the timestamp of a late gesture, disappearing prop or changing face. This gives you a concrete basis for deciding whether the clip is useful or needs another attempt.
If you revise, change the instruction connected to the problem rather than rewriting every part of a working prompt. Preserve the input and settings you want to compare. Generation is variable, so a changed output is not proof that one prompt edit caused every improvement. Keep the versions and task references together. Quality concerns and technical failures have different support paths; neither should be described as a guaranteed refund without reviewing the applicable policy.
Download and share responsibly
Use the result’s download action and play the saved file through to the end. Confirm it opens on the device or application where you intend to use it. Check the soundtrack and ending again after download rather than relying on the browser thumbnail. Keep your own copy, because hosted history is not promised as a permanent archive.
Before publishing, review rights in every input and any music, marks or likenesses present in the result. Do not describe synthetic footage as evidence of an actual event. Clearly label manipulated portrayals and obtain any additional permissions your use requires. For a support request, include the task reference and affected timestamp; for an access or deletion request, use the privacy contact. Do not send passwords, card details or unnecessary personal documents.
Burning Bridges party tool · AI image workspace · Subscription pricing · Support and privacy contact · Acceptable use