Protoface + Seedance 2.0 Production Test — What We Learned from a 15-Second Multi-Scene Video

FourFeetz Studios has recently been using Protoface Studio and Seedance 2.0 as part of our practical AI pet-video production workflow.

For this test, we moved beyond generating a single action and explored how several short memories could be connected within an approximately 15-second sequence.

The pet-memory film begins with a peaceful moment at home, transitions into the dog running happily outdoors, and ends in a warm golden-hour meadow.

One of the most interesting results was how multiple environments could still feel connected when they shared the same emotional direction.

However, we also learned that a longer prompt does not automatically produce better control.

What mattered more was simplifying the main action in each scene.

At home: rest comfortably beside a favorite toy.

Outside: run happily through the grass.

In the final meadow: remain peaceful, look toward the camera, and gently wag the tail.

Giving each section one clear primary action made the overall sequence easier to manage.

Negative direction was also important.

For this memorial project, we explicitly asked the model not to have the dog walk away, disappear, or create imagery associated with death or farewell.

When a film depends on a specific emotional tone, defining what should not happen can be almost as important as describing what should happen.

Our practical takeaway from this production was simple:

Give each scene one primary action.

Use only the camera movement you actually need.

Describe the pet's behavior clearly.

Explicitly exclude unwanted emotional or visual outcomes.

And for important titles or memorial messages, adding typography during final editing can be more reliable than asking the video model to generate the text itself.

Through these experiments, FourFeetz Studios is focusing not only on AI generation quality, but on building a production workflow that can be used for real creative services.

The supplied hotel series presents arrival, resting awake, running on grass, and sleeping as four separate states. These are AI concept images from another project, not extracted frames from the approximately 15-second test discussed above. They provide concrete comparisons for the size of an action assigned to each shot.

In the arrival image, the threshold and lobby establish direction. In the running image, the puppy approaches the camera with paws visible against the grass. The bed and sleeping images depend more on posture and expression. Keeping those different roles avoids adding travel and camera motion to every shot.

The awake and sleeping images both include bedding and a toy, but that does not establish an identical room. Before joining such shots, compare the room layout and prop positions; mismatched images can work as separate memories instead. Logos and lettering visible in the stills also need a readability check in the final edit, since their stability in motion is not established by these images.