May 8

How to film a POV vlog with a giant megalodon in the ice using Seedance 2.0



I did everything in SYNTX.AI

The AI models I used were Seedance 2.0 Pro and Chat GPT Image 2.0.

1. Preparing photos for the project

  • Upload your face photos in a 3x3 grid on a white background, showing different angles. (Important: This is the only way to generate a video using your own face.) [Photo Placeholder: Your 3x3 face grid]
  • Generate a single photo of a frozen megalodon using Chat GPT Image 2.0. Prompt for the shark generation:

Breathtaking yet terrifying cinematic arctic landscape, heavy overcast storm clouds, oppressive gloomy sky, swirling snowstorm with low visibility. A massive jagged glacier cliff rises like a vertical wall of ancient ice from a pitch-black frozen ocean. The sea below is violent, dark, and almost lifeless, with fractured ice chunks drifting like debris in a graveyard. Inside the semi-transparent glacier wall, a colossal prehistoric megalodon shark is frozen — but it feels disturbingly alive. The ice does not simply contain it, it suffocates it. The creature is shown in a dominant 3/4 profile, its gigantic head and upper body filling nearly the entire internal volume of the glacier, creating an overwhelming sense of impossible scale. The proportions feel unnatural and nightmarish, as if the glacier grew around a monster too large for the world. Its eye is faintly visible through multiple layers of ice — not fully clear, but unsettlingly present, as if it is still aware. The massive jaws are partially exposed through fractured ice planes, distorted by refraction, revealing rows of enormous teeth embedded in frozen depth. The mouth appears slightly open, frozen mid-motion, giving the illusion of a silent, trapped roar. Cracks radiate through the glacier like veins, subtly framing the megalodon’s skull and spine beneath the ice. Air pockets and frost distort its silhouette, making it shift visually depending on angle, as if it might move when not directly observed. The ice feels fragile in places, like it is barely containing something prehistoric and impossible. The glacier dominates everything — towering, vertical, oppressive — with the shark appearing to fill most of its interior like a living nightmare encased in stone-cold time. The environment feels wrong, silent, and tense, as if sound itself has been swallowed. Cold volumetric fog, extreme atmospheric perspective, low contrast shadows, desaturated blue-gray tones, minimal light breaking through clouds. Subtle distortion in ice refraction enhances unease. Photorealistic, ultra-detailed, cinematic horror realism, 35mm lens, shallow depth of field in foreground ice, subtle film grain, documentary-grade realism mixed with psychological horror tone, high dynamic range. --ar 9:16 --no people

Reference Video: We need this to define the camera movement so the AI understands exactly how the frame should shift. (Film yourself, then smoothly pan the camera forward and tilt it upward).

IMPORTANT: No human faces should be visible in the reference video. You can blur them out, film in a way that hides the face from the start, or go the easy route: wrap a scarf around your face like I did and film it that way! XD


Step 2: Generation in Seedance 2.0 (Settings and Prompt)

The entire assembly was done in Syntx.ai. Use the following settings: Video section -> Seedance 2.0 -> Pro Mode -> 720p -> Aspect Ratio 9:16 -> Duration: 7 seconds. Upload all three prepared files in order.

🔥 CRUCIAL STEP: When you copy and paste my prompt into the AI window, the system won't automatically link the files to the text. You must manually delete each tag (@Image1, @Image2, @Video2) in the prompt, type the "@" symbol again, and select the uploaded file from the drop-down menu. This is the only way to ensure the files are correctly linked for generation!

VIDEO PROMPT: Before starting, make sure the tags correspond to your uploaded files.

Total: 7s / Single continuous tracking shot / 9:16 aspect ratio. Style: Raw ungraded bystander phone footage, shot in auto mode, amateur handheld phone footage, natural imperfections, constant visible micro-shake, 35mm normal lens. Match the exact misty atmosphere and deep blue tones of @Image2. Overcast gloomy weather, light snow blowing horizontally.Reference @Video2 strictly for all camera movements, panning, pacing, and character choreography. Replace the person in @Video2 with the exact facial identity from @Image1. The person from @Image1 is wearing a heavy, fitted dark winter jacket with NO hood, head completely uncovered. Replace the original background completely with the towering, sheer ice wall from @Image2.Handheld selfie POV. [00:00-00:01s] A young man (@Image1) wearing a heavy fitted dark winter jacket (no hood, hair visible) stands on a massive, completely stationary, solid ice sheet. He looks slightly up and off-camera with intense, locked focus and frozen stoic awe. Visible cold breath vapor rises from the man. Natural handheld sway inherited from @Video2. [00:01s-00:07s] The camera executes a simple, smooth pan away from the subject to face forward and tilts up (matching @Video2), leaving the man entirely out of the frame. The view shifts exclusively to a stable forward and upward view, 30 meters away in the far distance, revealing the exact towering, sheer ice wall rising from the dark freezing water. Background Depth: On the left side of the towering ice wall, a vast, infinite dark freezing ocean stretches far into the deep horizon. The ocean must have visible depth extending far behind the ice wall, with scattered floating ice chunks. Frozen deep within the massive, semi-opaque blue ice is the colossal head of a terrifying megalodon, its giant open jaws and massive teeth looming through the frost, precisely replicating @Image2.Constraints: Weight identity reference @Image1 at 70% and motion reference @Video2 at 30%. Man MUST leave the frame completely after the pan. Preserve the exact architectural geometry of @Image2. Avoid identity drift, no hood, stationary ground, no fisheye lens distortion, no camera cuts, no scene transitions, no 3D, no cartoon, no VFX, no 180-degree camera rotation, no spinning, no flat fog backgrounds, ensure deep ocean horizon, maintain realistic physics. Native audio: howling freezing wind and distant ice cracking

Second Prompt.

Total: 7s / Single continuous tracking shot / 9:16 aspect ratio.
Style: Raw ungraded bystander phone footage, shot in auto mode, amateur handheld phone footage, natural imperfections, constant visible micro-shake, 35mm normal lens. Match the exact misty atmosphere and deep blue tones of @image2. Overcast gloomy weather, light snow blowing horizontally.

Reference @video1 strictly for all camera movements, panning, pacing, and character choreography. Replace the person in @video1 with the exact facial identity from @image1. The person from @image1 is wearing a heavy, fitted dark winter jacket with NO hood, head completely uncovered. Use @image2 ONLY as a style and color palette anchor.

Handheld selfie POV. A young man @image1 wearing a heavy fitted dark winter jacket (no hood, hair visible) stands on a massive, completely stationary, solid ice sheet in the middle of a dark freezing ocean, looking with intense, locked focus and frozen stoic awe. Visible cold breath vapor rises from the man. Natural handheld sway and micro-jitters inherited from @video1. At [00:02s], the camera smoothly but completely pans hard left, leaving the man entirely out of the frame. The view shifts exclusively to the empty dark water. In the far distance, 30 meters away, a mountain-sized, terrifying megalodon dorsal fin (three times the size of a whale) breaks the surface, swimming past and slowly submerging. Heavy water displacement and turbulent wake.

Constraints:Man MUST leave the frame completely after the pan. Avoid identity drift, no hood, stationary ground, no fisheye lens distortion, no wide-angle lens, no camera cuts, no scene transitions, no morphing, maintain realistic physics. Native audio: churning freezing water, heavy splashes, and low frequency rumble.




You can find more tutorials and helpful prompts like this on my Telegram channel:
[https://t.me/eyeaii]