Four-Model Photoreal Prompt Test

Prompt

1. A face and a spoken line

Prompt:
"Slow push-in on the woman leaning against the wall. She holds the camera's gaze, exhales softly through her nose, and a small knowing smile forms at one corner of her mouth. She says quietly: "I already said yes." Natural daylight, real skin texture, subtle handheld breathing in the camera. Quiet room tone."

Each pass plays a different model's audio. Seedance 2.0 held a closed-mouth smile and dropped the "I" from the line. The other three said the full line.

2. A group laughing

Prompt:
"Handheld camera, warm bar light. The group bursts out laughing together; the platinum-haired woman throws her head back laughing, and two friends beside her clink their glasses. People lean in, shoulders shake with laughter. Busy bar chatter, laughter, glasses clinking."

Each pass plays a different model's audio. Kling 4.0 and Seedance 2.0 show the clearest glass clink.

3. Hands handling small objects

Prompt:
"Static camera, low-key lighting. The man's fingers riffle the stack of poker chips, then slowly slide the whole stack forward across the table. His eyes lift to the camera, expression unreadable. Chips clicking softly, quiet room."

Each pass plays a different model's audio. Watch the chips: some models slide the stack forward, others scatter it.

4. Fire and smoke in 21:9

Prompt:
"Slow crane down into the smoke-filled city street. Flames ripple and flicker out of the building windows, glowing embers drift through the thick haze, and smoke rolls slowly between the towers. Dim headlights glow through the smoke at street level. Low roar of fire, distant sirens."

Each pass plays a different model's audio. All four kept the yellow car and the burning windows. None made the fire roar and sirens I asked for: H3, Kling and Seedance added music, and FLUX is nearly silent.

Prompt as published by the creator · https://x.com/HBCoop_/status/2113662571362590905

How to use this prompt

1. Copy the prompt above.
2. Open SunoMV and upload your song (Suno, MP3 or a link).
3. Paste the prompt as the visual style, then generate — tweak the scenes until it fits your track.