Use media you can authorize
Own the source or have permission to use it, including informed consent for a real person's likeness.
Papiboom Studio
GUIDE / REFERENCE-LED VIDEO
Connect authorized images and media to a compatible video model, direct each source clearly, and keep the selected result ready for the next step.
No subscription required / Pay-as-you-go credits
This is an illustrative test recipe, not a published generation result or a model-quality benchmark. Use your own licensed product photo, without people, readable private information or third-party branding.
Upload one clear photo of a plain fabric bag on a table. Select that exact result in the source node's history and connect it as a general image reference.
Select Seedance 2.5, 5 seconds, 720p, 16:9. Do not add first/last-frame roles or other references for this first trial.
Use the reference for the bag's shape, fabric and color. A locked-off camera shows the bag resting on the table while soft daylight moves across the fabric. Keep the bag stationary; introduce no new objects or text.
Compare silhouette, fabric, color, unwanted objects and camera movement against the source. Record the selected input, settings, task outcome and observed changes. A valid request does not prove those details will be preserved.
Change only one variable, such as the camera direction. Keep the source and other settings fixed. Each submitted generation can consume credits; check the displayed estimate before running.
If the request fails before output, use the input-check examples. For frame control use adaptive aspect and separate frame mode from general references. For model selection see Seedance 2.5 and the 2.5 versus 2.0 comparison. For permitted mature creative themes, read what Fewer restrictions does and does not change.
/ Papiboom editorial
A reference-led generation is not simply a prompt with an attachment. The media must finish uploading, remain connected, match a supported role, and reach a model that accepts the resulting input combination.
An AI video generator with reference images should preserve the relationship between source media, prompt direction, model settings, and the selected output. In Papiboom, connected references are submitted even without an @ token; the token is optional direction for how a specific source should be interpreted.
Own the source or have permission to use it, including informed consent for a real person's likeness.
Reference counts, media types, frame roles, duration, and combinations remain model-specific.
When a source node has several outputs, select the result that should flow downstream.
Choose the role before choosing the model. This prevents a valid file from being sent through an unsupported control.
Connect the image node and describe the source's intended role
Model inputsChoose frame mode and use one compatible image as the start
Model inputsProvide the first frame before adding the last frame
Model inputsUse a model or workflow with a dedicated motion-source input
Model inputsUse only when the selected model exposes audio reference support
Model inputsBegin with one known-good source. Add complexity only after the first submission path is working.
Upload or import an authorized image, video, or audio source and wait for its preview to become readable.
Attach the source to the generation node and confirm that its reference capsule appears.
Use a general reference, frame anchor, motion source, or audio input supported by the model.
Explain what should remain, change, move, or define the shot. Use @ when explicit source attribution helps.
Generate one or more takes, then select the exact history result that should feed the next node.
Duplicate or connect a downstream node while preserving the selected source and workflow context.
Short answers for common reference-led video decisions.
Yes when the selected model accepts multiple references. Papiboom keeps each source as a visible node and validates model-specific media counts and combinations before submission.
No. A general reference can guide subject, style, or scene context, while a first frame anchors the opening composition. Models expose these roles differently.
No. Papiboom submits a valid connected reference even without an @ token. Use the token when you want to explain the role of a specific source inside the prompt.
The selected result in that node's history is used downstream. This lets you compare iterations before deciding which image or video should guide the next generation.
Open Papiboom Studio, connect your references, and keep every iteration in context.