0:00 / 0:00

Phil Franco | AI Tools Expert 💰

Q: Why is the top panel important?

The top panel keeps the source image visible at all times, making it much easier to compare how different models interpret and extend the same prompt.

Q: What does the reel actually test?

The reel tests how different AI video models preserve the original character and scene while expanding the concept into larger cinematic environments.

Q: Why does the video end in space?

The Earth-orbit ending demonstrates the maximum scale of scene expansion, making model differences more obvious and visually dramatic.

@ai.withphil

INSTAGRAM · 2026-02-17Source

275likes

184comments

Remix This

Recreate with Seedance 2

Make your own AI viral video

Prompt

GLOBAL LOCK: A vertical AI model comparison reel with a fixed stacked layout. The top section shows the source image labeled "ONE IMAGE & PROMPT" and the bottom section shows the generated video output labeled by model, primarily "Seedance 2.0" and "Kling 3.0." The source image and generated scene are based on the same character: a young fair-skinned woman with dark shoulder-length hair, wearing a futuristic black jacket with bright pink accents, standing behind semi-transparent glass in a sleek sci-fi room while blowing a large pink bubble gum bubble. The words "get ready." appear on the glass. Keep this source identity and comparison layout stable while the generated output expands into larger sci-fi environments.

[00:00-00:06] Present the comparison layout clearly. The top remains the static source image and prompt frame. The lower generated output begins as a close interpretation of the original scene: the woman behind the glass, pink bubble gum inflating, dark hair, black-and-pink jacket, cool futuristic interior lighting, and the "get ready." text readable on glass.

[00:00:06-00:12] The generated output starts to diverge and expand. The bottom panel shows the camera pulling wider inside the same futuristic room, then pushing the scene outward into a larger metallic environment. The source image above stays unchanged as the control reference.

[00:00:12-00:18] The lower panel escalates into a space-station-like structure, with mechanical arms, orbital modules, or exterior sci-fi architecture replacing the interior room. The comparison demonstrates cinematic scene extension beyond the original portrait prompt.

[00:00:18-00:24] The generated output reaches a dramatic Earth-from-space view, with the planet curvature dominating the lower panel and orbital hardware or a station silhouette visible nearby. The top panel still shows the original bubble-gum portrait, reinforcing the scale jump from one image and prompt.

[00:00:24-00:30.2] The video cycles through variations of the expanded orbital sequence while model labels change between Seedance 2.0 and Kling 3.0. The final moments return attention to the comparison concept itself: how each model treats continuity, expansion, and visual ambition from the same input.

SUBJECT: one dark-haired young woman in a futuristic black-and-pink jacket blowing a pink bubble gum bubble behind glass, shown as a reference image and then interpreted by different AI video models.

ENVIRONMENT: stacked comparison layout on black background; initial sci-fi glass room; later expansion to futuristic corridor, space-station hardware, and Earth orbit.

ACTION: bubble gum close-up, room expansion, orbital reveal, Earth-from-space cinematic pullback, model-to-model comparison.

CAMERA: vertical 9:16 social explainer format with fixed top reference panel and dynamic lower generated panel; the generated panel uses cinematic pullbacks and world-expansion framing.

LIGHTING: cool blue-white futuristic lighting in the glass room, then darker space contrast with bright Earth highlights and metallic reflections.

GRADE: clean AI-tool explainer grade, crisp UI-like presentation, cool sci-fi palette with pink bubble gum accent, high contrast in space shots.

MOTION: subtle portrait motion at first, then major environmental expansion and scale transitions in the generated output panel.

SPEECH PACK: no subtitles visible in the comparison itself beyond labels; the reel functions as a visual demo/tutorial. No dialogue needs to be represented in the scene prompt.

NEGATIVE PROMPT: remove the comparison layout, generic talking-head explainer, unrelated stock footage, fantasy medieval settings, warm rustic tones, extra characters, cluttered interface overlays, night-club look, missing "get ready." text, no Earth-orbit escalation.

How ai.withphil Made Seedance Kling One Image Prompt Comparison — and How to Recreate It

This video is a clean comparison-format reel built around a simple question: how far can different AI video models extend the same starting image and prompt? The layout stays stable throughout. The top panel shows the fixed source frame labeled "ONE IMAGE & PROMPT," featuring a dark-haired young woman behind glass blowing a pink bubble gum bubble in a sleek futuristic room with the words "get ready." on the glass. The lower panel then shows model outputs labeled by system, primarily Seedance 2.0 and Kling 3.0. That fixed structure is what makes the reel useful. The audience always knows what the baseline is and what is being compared.

The real educational value comes from the scope of the expansion. The generated panels do not only animate the face or bubble. They widen the scene into a larger sci-fi interior, then push outward into mechanical orbital imagery and eventually a dramatic Earth-from-space composition. This makes the reel relevant for search intent like Seedance vs Kling comparison, one image to cinematic video prompt, AI scene expansion tutorial, bubble gum glass room prompt breakdown, and Earth orbit pullback AI video workflow.

What happens in the first 0 to 6 seconds

The first moments establish both the source character and the comparison logic. The top reference image is static and readable. The woman has dark shoulder-length hair, a black-and-pink futuristic jacket, and a large pink gum bubble. The lower generated panel initially stays close to that source setup, which helps viewers orient themselves. This matters because the later jump into orbital scale only makes sense if the audience first understands the original scene and prompt identity.

Comparison breakdown

00:00-00:06: The reel establishes the stacked comparison layout and shows the generated panel staying near the source portrait scene.

00:06-00:12: The bottom panel begins widening the environment, revealing more of the futuristic room and hinting at scene extension beyond the original crop.

00:12-00:18: The generated output moves into mechanical or orbital architecture, showing how the same prompt concept can scale into a much larger world.

00:18-00:24: Earth curvature and space imagery dominate the lower panel, creating the most dramatic contrast against the fixed portrait reference above.

00:24-00:30.2: The clip cycles through further comparison variations, reinforcing the tool-evaluation point rather than introducing a new concept.

Why the format works

The key strength here is methodological clarity. Many AI comparison videos jump between outputs without preserving the input context, which makes them hard to learn from. This reel solves that by locking the source image and prompt at the top. The viewer can always compare the generated panel against the same baseline. That makes the expansion from portrait to orbital scene feel measurable rather than arbitrary.

The source character also matters because it is visually strong enough to survive the comparison format. The pink gum bubble, black-and-pink jacket, and "get ready." text create a memorable anchor. That anchor helps viewers track whether the models preserve identity, styling, and prompt intent as the scene gets more ambitious. For creators, this is a strong example of how to design a source image that is both aesthetically compelling and useful for testing.

Prompt reconstruction notes

To rebuild a reel like this, think of it as two prompts layered together. The first prompt defines the comparison UI structure: stacked layout, fixed source panel, labeled model outputs, black background. The second prompt defines the cinematic source scene itself: the dark-haired woman, black-and-pink jacket, pink bubble gum bubble, futuristic glass room, and the phrase "get ready." on the glass. After those are locked, you describe the generated panel's ambition to expand the scene outward into larger sci-fi architecture and ultimately an orbital Earth view.

This clip is especially useful for teaching prompt scalability. It shows that one image and prompt can be used not only for micro animation but also for world extension. That makes it valuable for anyone studying camera pullbacks, scene continuity, or how different models handle narrative escalation from the same seed image.

How to rebuild this comparison format

Step 1: Create a visually distinctive source image with one clear character and one memorable action, such as bubble gum inflation behind glass.

Step 2: Lock a stable stacked comparison layout with the source image on top and the generated output panel below.

Step 3: Label the input and the model outputs clearly so the audience can evaluate differences without guessing.

Step 4: Start the generated panel close to the original portrait before widening or extending the environment.

Step 5: Use scene expansion as the test variable, pushing from the initial room into more ambitious architecture or orbital worldbuilding.

Step 6: End on the widest or most cinematic variation so the reel demonstrates the maximum scope difference across models.

Replaceable variables

You can swap the source character, action, or environment as long as the input frame remains visually distinctive and the generated outputs are asked to extend the same concept. You can also compare different pairs of models, but the fixed source panel should always remain on screen if the goal is educational clarity.

Common failure cases

The biggest miss is removing the fixed reference image, which turns the comparison into a loose montage rather than a useful test. Another is choosing a source image that is too generic to track across model outputs. A third issue is expanding the generated scene too abruptly without first showing a close interpretation of the original frame, which weakens the viewer's understanding of continuity.

Publishing and SEO growth actions

This page should target creators evaluating AI video tools and prompt extensibility. Strong query angles include Seedance 2.0 vs Kling 3.0 comparison, one image and prompt AI video test, prompt-consistency video workflow, and cinematic scene extension from a single source frame. The page becomes useful when it explains input stability, output escalation, and why a strong source image makes model differences easier to interpret.

FAQ

What is the fixed input in this comparison? The fixed input is a portrait of a dark-haired woman blowing a pink bubble gum bubble behind glass in a futuristic room, labeled as the one image and prompt reference.

Why is the top panel important? It keeps the source visible at all times, which makes the lower generated outputs easier to evaluate and compare.

What does the reel actually test? It tests how different AI video models preserve character identity and extend the scene outward into larger cinematic environments from the same starting prompt.

Why does the video end in space? The Earth-orbit sequence demonstrates the maximum scale of scene expansion, showing how far each model can push the same initial concept.