May 28

How to Create a “War of the Worlds” POV Vlog with Seedance 2.0

I made everything in SYNTX.AI

I used Seedance 2.0 Pro, ChatGPT Image 2.0, and a few secret tools too — but more on that at the end :)

In this guide, we’ll break down only the first segment of the video.

1. Preparing the photos for the project

  • Place your face photos in a 3x3 grid on a white background, using different angles.

This is important because it helps the AI keep your face consistent in the video and prevents it from changing your appearance during camera movement.

  • Next, we need to prepare an image of the tripod that will be used as a reference for the video.

To do this, open ChatGPT Image 2.0 and ask it to create a character sheet of the tripod based on a movie frame where the shape of the alien machine is clearly visible.

Important: we don’t need just a beautiful image. We need a clear reference for the AI, so it can understand the silhouette, head shape, legs, cables, searchlights, and overall tripod design.

Example prompt:

Create a character sheet of a giant alien tripod war machine based on the visual reference from the uploaded movie frame. Show the tripod clearly from multiple angles: front view, side view, and 3/4 view. Preserve the main silhouette, head shape, three long mechanical legs, hanging cables, searchlight-like glowing eyes, and dark metallic surface.The sheet should be clean and readable, with the tripod separated from the background, neutral lighting, full body visible, no cinematic scene, no smoke covering the design, no people, no destruction. The goal is to create a clear reference sheet for video generation.

  • Now we need a reference video that will be used as an example for the camera movement.

Record a simple video: start with a selfie-style shot, then smoothly turn the phone toward the direction where the street, the running crowd, and the tripods will appear in the final generation.

IMPORTANT: your face should not be visible in the reference video, otherwise the generation may not pass moderation.

This video is needed only for camera movement, pacing, turning motion, and framing behavior. The background from this video will be completely replaced.


Step 2: Generating the video in Seedance 2.0 — Settings and Prompt

I made the whole setup in Syntx.ai.

Settings:

Video section -> Seedance 2.0 -> Pro mode -> 720p -> Aspect ratio 9:16 -> Duration 9 seconds.

Upload all 3 prepared files in this order:

  1. @Image1 — 3x3 face grid
  2. @Image2 — tripod character sheet
  3. @Video1 — reference video with camera movement

VERY IMPORTANT: When you copy my prompt below and paste it into the AI prompt field, the platform may not always attach the files to the text automatically. Make sure every tag correctly matches the right attachment.

VIDEO PROMPT:

Total: 9s / Single continuous handheld smartphone shot / 9:16 aspect ratio.Style: raw ungraded bystander phone footage, amateur handheld smartphone camera, natural imperfections, realistic phone optics, 35mm normal lens, no stylization. Overcast weather, dense fog, bleak Russian urban atmosphere, grounded realism.Use @Video1 strictly for camera movement, pacing, framing behavior, and choreography.Use @Image1 strictly for the exact facial identity of the man.Use @Image2 strictly for the tripod design, silhouette, head shape, leg structure, glowing searchlight eyes, cables, and appendages.Replace the original environment completely with an ordinary Russian urban street: roadway, sidewalks, muted residential buildings, bare trees, damp asphalt, dense fog, and everyday city surroundings. No winter, no snow.[0-3s]Handheld selfie POV. A young man from @Image1 stands on the Russian street wearing a dark heavy coat. He does NOT look into the camera. He is already looking past the phone toward the exact off-screen direction that will be revealed next, as if he has noticed something enormous there. His expression is tense, uneasy, anxious, and stunned — quietly shocked and alert, but not screaming and not overly panicked. Natural handheld sway and micro-jitters inherited from @Video1. He does not speak.Many civilians are already running in panic nearby, crossing through the background and the edges of the frame. One fleeing civilian lightly bumps into the man’s shoulder in passing. The contact is brief and accidental. The man barely reacts and keeps looking forward toward the off-screen threat without breaking focus.Important crowd rule for this section: all visible civilians must be fleeing in panic. Nobody is standing still, walking calmly, hesitating, or watching passively.[3-9s]The camera smoothly but shakily turns away from the man's face, matching @Video1 exactly, and the man leaves the frame completely. The view reveals the foggy street, residential buildings, and a larger crowd of civilians running away through the street and sidewalks.Crowd behavior rule:There must be many people visible, and every visible civilian is actively fleeing in panic. All of them are running, sprinting, stumbling, or rushing away from the threat. No one is standing still. No one is calmly walking. No one is frozen in place. No one is casually observing. The crowd must clearly read as a panicked evacuation.Exactly 2 alien tripods are visible, matching @Image2 exactly.ABSOLUTE SCALE LOCK:Both tripods must be ENORMOUS and clearly far taller than the buildings. Each tripod is at least 2 times taller than the nearby buildings. They must tower high above the rooftops and feel truly threatening. They must NOT look like small robots, medium-sized walkers, or tiny distant figures. The surrounding buildings must look small beneath them.Tripod 1:The first tripod is the nearest and main threat. It is positioned behind a nearby building, and the building must hide almost all of its lower structure. Only its enormous upper body, head, uppermost leg sections, glowing searchlights, and hanging cables are visible above and behind the roofline. Its body is also partially obscured by dense fog. The rest of its legs, lower body, and feet are completely hidden behind the building and fog.Important: do NOT show the lower legs or feet of the near tripod at all. Do NOT animate visible stepping for the near tripod. Its lower structure remains concealed. The visible upper section must still tower far above the roof, proving that it is gigantic.The near tripod does not walk toward the viewer. Instead, it slowly turns its head and upper body toward the camera, as if noticing the observer. This turning motion must feel slow, graceful, eerie, deliberate, and predatory.Tripod 2:The second tripod is much farther away, visible down the road in the distance. It is also gigantic, clearly taller than the surrounding buildings, and partially obscured by fog. This is the tripod that is visibly moving. It walks slowly along the road in another direction, NOT toward the camera, NOT toward the observer.MOTION RULE:Only the distant tripod should have clearly visible walking motion. Its movement must be slow, graceful, elegant, and deeply unsettling — long controlled steps, smooth weight shifts, subtle body sway, no jerky animation, no awkward robotic walking, no clumsy motion.The near tripod should not show visible full-leg walking. It should remain mostly concealed and only perform a slow upper-body/head turning motion.SPATIAL RULE:The tripods must NOT move toward the camera. They must NOT float, hover, or step through empty air. The first tripod stays partly hidden behind the building and partly hidden by fog while turning toward the observer.The second tripod remains far away, partly hidden by fog, and continues moving in another direction along the road.Audio: diegetic sounds only — heavy tense breathing, distant screams, many panicked footsteps, scattered people running, distant sirens, and a deep terrifying mechanical horn echoing from the tripods. No dialogue, no talking, no yelling by the main character.Negative constraints:no looking into camera, no speaking, no tiny tripods, no small robots, no toy scale, no floating tripods, no hovering, no full-body comfortably framed miniature tripod figures, no visible lower legs or feet on the near tripod, no visible stepping animation on the near tripod, no civilians standing still, no civilians calmly walking, no passive bystanders, no fast tripod movement, no jerky motion, no attack toward the viewer, no CGI look, no cartoon, no 3D, no fisheye, no camera cuts, no scene transitions, no drone shot.

In this guide, I showed how to create the first segment. I’ll break down the full two-part video setup, as well as how to achieve this level of realism, separately in the full guide inside my bot, which you can find in the pinned message of my Telegram channel.
https://t.me/eyeaii