Why Seedance 2.0 reference?

It's the strongest video model available right now for holding a consistent look across multiple scenes. Other models drift on lighting and detail between shots. Seedance keeps the world stable so the scenes feel like one piece.

What's the most important part, the model or the prompt?

The prompt. The model is good, but if you leave the prompt open-ended, AI fills the gaps with things you didn't ask for. The creators getting clean results are the ones writing exact, directed prompts.

How do I "timestamp" a shot?

Tell the model what happens at each moment: "0-2s wide establishing shot, 2-4s push in on the product, 4-6s tight detail." Treating it like a shot list instead of a vague description is what makes scenes land.

Why does it look different every time I run it?

That's expected. The same prompt produces different takes. Use it. Generate five or six and pick the best one. The version you see in the video was take six.

What do I do with the final clip?

Trim and cut it in whatever editor you use (CapCut, Premiere). Generation happens in Ophelia, final assembly happens in your editor.

Ophelia Guide

Make a Viral AI shot from one image, Guide in Ophelia:

Turn a single photo into a cinematic multi-scene shot with Seedance 2.0. This free Ophelia guide breaks down the exact prompt structure and rerun process behind the viral version.

Start with one image. A product photo, a screenshot, any single frame
Generate your starting frame in Ophelia. This sets the look for the entire shot
Drop your starting frame into the main prompt box
Toggle Seedance 2.0 Reference (the strongest model for this right now)
Paste the prompt below and write your shot list inside it. Be exact, AI fills open gaps with things you don't want
Hit generate, then rerun until it's right. The good version is usually take five or six, not take one

The prompt

"Using image @1 as the style reference, create a cinematic multi-scene shot built from this single frame. Hold the exact look of image @1 throughout: same color grade, contrast, lighting, and overall aesthetic, so every scene feels like one continuous piece.
Direct the sequence like a shot list. Describe what happens and when, move between distinct camera angles, and keep the world and subject consistent from scene to scene.

[describe your scene shot by shot here. say what happens at each moment and timestamp the shots, e.g. 0-2s wide establishing, 2-4s push in on the subject, 4-6s detail shot. the more exact, the closer it lands]

Keep the motion smooth and motivated, the pacing cinematic, and every frame clean. No flicker, no warping, no text or logos unless specified.
Output a high-resolution cinematic clip, ready to trim and cut in post."

Importanttip: you can hand this prompt to Claude and talk through your idea. it'll help you structure the shots and timestamps based on what you're making.
Importantthe real secret: rerun. the same prompt gives different results each time. generate it five or six times and pick the best one. that's how the viral version gets made.

FAQ

Why Seedance 2.0 reference?: It's the strongest video model available right now for holding a consistent look across multiple scenes. Other models drift on lighting and detail between shots. Seedance keeps the world stable so the scenes feel like one piece.
What's the most important part, the model or the prompt?: The prompt. The model is good, but if you leave the prompt open-ended, AI fills the gaps with things you didn't ask for. The creators getting clean results are the ones writing exact, directed prompts.
How do I "timestamp" a shot?: Tell the model what happens at each moment: "0-2s wide establishing shot, 2-4s push in on the product, 4-6s tight detail." Treating it like a shot list instead of a vague description is what makes scenes land.
Why does it look different every time I run it?: That's expected. The same prompt produces different takes. Use it. Generate five or six and pick the best one. The version you see in the video was take six.
What do I do with the final clip?: Trim and cut it in whatever editor you use (CapCut, Premiere). Generation happens in Ophelia, final assembly happens in your editor.