Motion Transfer Reference Video: A Two-Input Checklist
Choose your source media and review the settings before processing.
Open the editing toolA photo defines what should appear. The driving video supplies the performance to follow. A detailed prompt cannot replace the missing performance file. If all you have is a photo and you want invented movement, start with image-to-video instead.
Open Motion Transfer when you have both files ready. It lives on the Edit Video page: the Motion Video (drives the animation) field takes the driver and the Character Image field takes the photo. Opening the tool does not generate a result or charge credits.
Check the actual input requirements first
Viralance uses DreamActor v2 for this workflow. The fal input specification requires an image URL and a video URL. It lists JPEG/PNG reference images up to 4.7 MB and MP4, MOV or WebM driving videos up to 30 seconds, with resolution bounds for each. Viralance's own uploader caps the driving video at 30 seconds and accepts common image files for the character. A provider's supported format does not mean every option appears in Viralance, so read any upload error before retrying.
The provider notes that the raw output starts with a roughly one-second transition, and its API trims that first second by default. Leave room around the important action and inspect the finished clip's timing; do not build an exact spoken cue around the very first second of the driver.
Match what the image shows to what the performance needs
For your first test, use a paired preparation sheet:
| Intended movement | Character photograph | Driving performance |
|---|---|---|
| A nod or small head turn | Clear face and shoulders | Face stays visible in a similar crop |
| A hand gesture | Hands and arms are inside the frame | Gesture stays away from the frame edges |
| A body movement | Enough of the body is visible | Subject remains visible throughout |
These are practical starting choices, not measured success rates. A portrait that ends at the chest gives the model little visual information about legs. A driver repeatedly turning away makes it harder to inspect the face match. Start with a simpler pair before attempting demanding choreography.
Use images and performances you are allowed to use. Pick your own character or an authorized subject; a successful upload does not establish permission.
Record a diagnostic driver
A useful first driver has a beginning, one action and a settled ending. For example, stand in a stable position, make one slow gesture, and return to that position. Keep the camera steady enough to distinguish the performer's movement from camera motion.
Watch the whole file before uploading. Look for an arm leaving the frame, a hand covering the face, another person crossing in front, a hard cut or a moment of blur. Write down the time of each problem. You now have a way to compare the output with the driver instead of guessing why an artifact appeared.
This short diagnostic is a suggested production exercise. We have not measured a universal framing or movement threshold that guarantees success.
Know what you are spending
Motion Transfer in the Viralance dashboard costs a flat 12 credits per run, checked September 26, 2026 — the same whether the driver is 5 or 30 seconds. That is about $3.60 at the Mini pack rate and about $2.17 at the Business rate. Review the displayed total before pressing Transfer Motion.
Because the price doesn't change with length, a short diagnostic driver saves no credits; its value is that it tells you faster whether the input pair works. The 4 free credits new accounts receive after verification do not cover a run; the $9 Mini pack (30 credits) covers two. Failed runs are refunded automatically; results you dislike are not. See the motion transfer overview for the broader feature and credit packs for paid options.
Diagnose the output before changing the model
| Symptom | First thing to inspect | A focused next change |
|---|---|---|
| A hand appears distorted | Driver hand position and occlusion | Use a slower, unobstructed gesture |
| Face detail drifts | Source face clarity and driver head angle | Use a clearer source or smaller turn |
| Body shape changes | Whether both inputs show enough body | Match their framing more closely |
| Action seems mistimed | Driver lead-in and output trim | Leave more room before the key action |
| Scene jumps | Cuts or camera motion in the driver | Try one continuous recording |
The table provides hypotheses, not automatic diagnoses. Keep the source image fixed when testing a different driver. Keep the driver fixed when testing a new photograph. Changing both at once makes it harder to identify what helped.
Frequently asked questions
Why is Generate unavailable after I upload a photo? Check that the driving video is also present, both uploads have completed, and your balance covers the displayed cost. Read any input error before retrying.
Does the result inherit perfect lip sync and identity? No. The model attempts to transfer performance. Watch the result closely and listen separately; movement, likeness and audio all need review.
Should I use the longest driving video first? The price is the same, but start with a short diagnostic when you are uncertain about the input pair. It is quicker to review, and problems are easier to trace to a specific moment in the driver.
