A greeting video for one person usually starts as a template with the name swapped in. This clip goes the other way: a kitten tears through a sheet of daisy paper, waves at the camera, and holds up a heart-shaped pillow with a message on it, generated in CapCut Video Studio at 16:9 in one ten-second take. The article covers the prompt, the settings, what the clip cost at those settings, how the stitched message behaved across two runs, and how to change the message in the editor without paying for another generation.
The finished clip, shown here as frames in sequence: the tear, the kitten, the wave, and the pillow held up to read.
Ten seconds, three actions, one recipient
The format is a small piece of staging. A flat surface faces the camera, something breaks through it, greets the viewer, and hands over a gift. The gift carries a message, so the same clip can be addressed to one person, which is why this format circulates as a birthday, thank-you, or holiday greeting rather than as a generic pet video. That gives the prompt three jobs in a fixed order: the tear, the greeting, and the hand-over, with the last one held long enough to read.
A paper wall is a landscape shot
The staging suits a 16:9 frame. The paper is the entire background, so a wide frame reads as a wall instead of a strip, the tear can sit in the exact center, and the kitten has room on both sides to climb out without touching the edges. The camera stays fixed and straight on. Any camera move would show that the wall is a surface with nothing behind it, and a fixed camera also keeps the message in one place for the final hold.
Write the actions in order and pin the camera
Fixed camera, straight on, landscape 16:9 frame. The whole frame is a flat sheet of cream paper printed with small white daisies, evenly lit, filling the frame edge to edge like a wall. First, a small tear opens in the center of the paper and a pink kitten nose pushes through; the paper rips wider and a fluffy orange tabby kitten with big round eyes climbs halfway out of the hole toward the camera, torn paper edges curling around its shoulders. Then the kitten looks straight into the camera and slowly waves one front paw, twice. Finally the kitten reaches back into the hole, pulls out a small red heart-shaped plush pillow with the words 'for you' stitched on it in white, holds it up toward the camera with both paws, and stays still looking at the camera for the last two seconds. The camera never moves. Soft realistic fur, shallow depth of field, warm gentle light, nothing else in the frame.
The order words carry the timing: First, Then, and Finally tell the clip what happens in which sequence, and stays still looking at the camera for the last two seconds reserves the end of the clip for reading the message. The camera never moves appears twice on purpose, once as the opening instruction and once at the end. torn paper edges curling around its shoulders ties the kitten to the surface so it reads as coming through the paper rather than standing in front of it, and nothing else in the frame keeps the wall clean.
Set 16:9 and 10 seconds before you send
In the CapCut AI video generator composer, the mode menu next to the plus button was set to Video. That turns the chip beside it into a settings row, which on 4 September 2026 read Seedance 2.0 Mini, Auto, 5s, 720p, and opens a panel with two tabs, Video clip and Full video. Video clip is described in the panel as generating a single-scene clip shorter than 15 seconds. Below that the panel offers the model and quality, an aspect ratio row running Auto, 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16, and durations of 5, 8, 10, 12, and 15 seconds. This clip used 16:9 and 10s, and the chip read Seedance 2.0 Mini, 16:9, 10s, 720p before the prompt went out.
The settings panel under the composer. Aspect ratio and duration are chosen here, before sending, and the chip above confirms the choice.
What one clip cost at these settings, and what came back
After sending, the workspace opened a canvas with the chat on the right, and the chat stated the price before generating: Generating 1 item will consume 80 credits. That figure applies to this configuration only, Video Studio in Video clip mode, Seedance 2.0 Mini, 16:9, 10 seconds, 720p, observed on 4 September 2026 while the Video Studio home page was showing a credit discount banner for Seedance 2.0 Fast and Mini. Other models, durations, resolutions, and modes are priced differently, and the chat shows the figure for whatever is set at the moment you send, so read that line rather than assuming a rate. The credit balance in the header read 2,179 before this clip and 2,099 after it, which matches the stated 80.
The chat states the settings it will use and the credits it will take before the clip is generated.
In this run the clip reached 100% after about four minutes, as a 10-second 16:9 video with its own audio track. The button under the card, labeled Export on hover, saved a 1280 by 720 MP4.
Four frames from the run at 0.6, 3.2, 5.2, and 9.2 seconds. Each action in the prompt shows up in the order it was written, and the final hold is long enough to read the pillow.
Two runs of stitched text
Generated text is the part of this format that most often goes wrong, so it was tested twice. The first run asked for for you. The second run used an otherwise identical prompt and asked for Happy Birthday, Maya. In both runs the words came out as legible white stitching with the capitals and the comma in the right places, and the second run broke the longer message onto three lines by itself. Two runs are two runs, and a third could still misspell a name. What they show clearly is the price of getting a message this way: every new name is a new generation at the full clip cost, and a misspelling means paying it again.
The two runs side by side, with the pillows enlarged below. Both messages were readable at full size in these runs.
Personalize in the editor instead of regenerating
The cheaper way to address the clip to someone is to keep the generated pillow as it is and add the name as a text layer in the web editor. The clip used here is the first run, with its stitched for you left in place. In the editor, the exported MP4 was dropped onto the preview area, which placed it on the timeline, and the Text panel in the left rail offers Add heading and Add body text under Basic. Add heading put a text clip on the track above the video and opened its settings on the right: a content box, a font menu reading Default, a size field, Style options for fill, stroke, background, and shadow, and a Transform group with scale and X and Y position fields. The content was replaced with Happy Birthday, Maya, the size was reduced until the line fit inside the frame, and Y was set to 250 to move the line to the top, clear of the kitten and the pillow. Dragging the text clip along the track put it over the last three seconds, so the name appears as the pillow comes up.
The name as a text layer over the generated clip, positioned with the Y field in the Basic panel. The pillow still reads for you, and the credit balance in the editor header did not move while the layer was added and the project exported.
Export, top right, opens a menu whose last item is Download, and Download opens an Export settings panel with a video cover, a name field, and menus that read 720p, Recommended quality, 30fps, and MP4 in this session. The Export button at the bottom of that panel produced a 1280 by 720 MP4 in about 25 seconds, with the name rendered into the frame.
The exported frame. The pillow is the generated part; the line at the top is the editor's text layer.
One thing went wrong in this pass and is easy to avoid. The text clip had been dragged so that it ended slightly after the video, and the export inherited that length: 10.37 seconds instead of 10.04, with about a quarter of a second of black at the end. Trim the text clip to the end of the video clip before exporting.
What to check before you send it
Read the message at full size, not in the thumbnail. Confirm the final hold: the pillow should stay in place and in focus through the last two seconds, and if the kitten lowers it early, the clip needs another run. Check the first second, because the tear has to be visible before a viewer scrolls on. Both generated clips arrived with an audio track, so listen before sending; the editor's left rail has an Audio panel, and a text layer only adds to the picture, so neither changes what the kitten does. If you exported from the editor, play the last second and make sure it ends on the kitten rather than on black.
Written 4 September 2026. The kitten, the paper, and the pillow are AI-generated in CapCut Video Studio; no real animal or product is depicted. Interface labels, timings, and credit figures reflect a single session on that date and the configuration named above, and may change.