TL;DR: This workflow turns a single AI-generated room into a continuous, never-ending cinematic descent by chaining image generation and image-to-video motion. 1. Core Idea Instead of generating separate scenes, you build a connected spatial system: - One base environment (the “anchor room”) - Infinite variations of that same space - Consistent camera motion across all clips This creates continuity that feels like a single endless world. 2. Step-by-Step Workflow Step 1 — Base Scene (Midjourney)Generate a single backrooms-style room: - mono-yellow walls - fluorescent lighting - drop ceiling - liminal, empty composition This becomes your reference “world state.” Step 2 — Scene Expansion (Nano Banana / remix tool)Instead of creating new environments, you: - remix the same room - slightly vary layout, depth, or structure - maintain visual continuity This prevents “world resets” between shots. Step 3 — Dual Image-to-Video Clips (per room) For each room, generate TWO clips: - Clip A: black frame → roomPrompt:“Camera descends vertically downward at a constant moderate speed.” - Clip B: room → continuationSame motion prompt, but starting from the room image. 3. The Chain You connect clips like this: Clip A → Clip B → next room → repeat This produces: - continuous camera motion - no visual resets - the illusion of infinite architecture Let me know if this help for if there are any questions!

TL;DR: The hardest part of AI video is camera control. "FPV drone shot" sends the camera anywhere but where you want. The fix is to literally draw the flight path as a red line on an aerial shot. Agent One on invideo draws that line for you, then one prompt tells it to remove the line and fly an FPV drone along the exact path. One image, one prompt, one unbroken cinematic shot. The problem Camera movement is what separates a real shot from a generated one, and it's the thing AI video is worst at. You type "FPV drone shot, fast, cinematic" and the camera drifts off doing its own thing. You can't tell it the path, so you settle for slow push-ins and pull-backs. The fix: draw the path Take an aerial or high angle still of your scene and draw a single red line across it. That line is the camera's flight path. Then in Agent One you paste a prompt that says: remove the red line, create a 15 second ultra fast FPV shot in one unbroken take that faithfully follows the path, and use the image as the full reference for everything in frame. The part that makes it effortless You don't even have to draw the line by hand. Hand Agent One the aerial shot and ask it to mark the flight path, and it returns the same image with the red line on it. From there it's one paste and you have a continuous drone shot diving through the scene exactly where you wanted it.

TL;DR: AI footage doesn't feel cinematic because of one perfect shot. it feels cinematic because of the contrast between shots* All calm bores; all chaos turns to mush. The fix is a rhythm — calm, chaos, calm — and since one model can't nail both, you split the job: **Kling for the calm beats, Seedance for the chaos beat.** --- The trap Everyone optimizes the single clip. They re-roll the same prompt chasing one flawless shot, then string a bunch of those together. The result feels flat — either uniformly smooth (and staged) or wall-to-wall chaos (and exhausting). Neither reads as cinema. The reframe The cinematic feeling isn't in a shot. It's in the contrast between shots. A perfectly stable shot feels staged. A run of chaotic shots becomes visual mush. What your eye actually reads as "cinematic" is the whiplash when you cut from one to the other. So you sequence it: calm → chaos → calm. Bookend your loud, high-energy beat with two quiet, controlled ones. The contrast does the work. Why two tools Beyond quality, it's economics: there's no sense burning your best video credits on a shot that's supposed to be still. So split the job: Kling — the calm beats. Holds composition, stays stable and controlled. Use it for the quiet bookends: a slow hover, a held wide, a clean push-in. Seedance — the chaos beat. Leans into motion, energy and impact. Use it for the one loud moment in the middle: ignition, action, things breaking loose. Kling → Seedance → Kling. Calm, chaos, calm. The best shots never come from one tool. They come from how you sequence them.

Welcome. I'm Jordano one of the Directors at DreamCodeFilms. This community exists because most AI video looks the same, same prompts, same outputs, no point of view. We do the opposite. Here we direct AI, we don't just generate with it. The craft comes first; the tools serve the story. So before anything else — introduce yourself in the comments. Tell us three things: 👋 Who you are (and where you're posting from) 🎬 What you're making right now, the project or style you're into 🎯 The one thing you want to be able to create by the time you leave here Everyone in this room started exactly where you are. Drop your intro below and say hi — and reply to one other person while you're here. That's how this place stays alive. Stuck or just want to talk shop? DM me anytime. I'm here. — Jordano

This is one of the videos I made on TapNow. My content focuses on the concept of film cuts, with more emphasis on dialogue. Some friends have given feedback on the lighting and color grading. What do you think? What could be tweaked to make it better?