How to direct AI video with 3D Builder previs

TL;DR

Directing first, then generating saves credits and keeps scene consistency. The 3D Builder tool auto blocks and labels scene elements, places a character and cameras, and lets you adjust motion, props, and cached prompts before rendering. You can export snapshots or videos and revise easily within the canvas.

Transcript

Top View just released a new tool that fixes the feel of being in a lottery. [music] Every time you generate AI scenes, it's called 3D Builder. Hi, I'm Julia. Here's what's happening every time you generate an AI video. You type a prompt and the model guesses. Sometimes it nails it. More often, it runs the wrong direction or the camera floats somew... Read More

Key Insights

• X is a cheap previs studio for AI video, enabling the director to specify staging, camera, motion, and timing before generation.
• The key to reducing lottery like results is directing first, then letting the AI render, which preserves consistency across shots.
• The tool auto labels all elements in the scene when prompted in detail, aiding revision and accuracy during blocking and export prep.
• Y is the ability to place multiple cameras and use presets or custom angles, then apply follow or look at targets to create dynamic shots.
• Inbuilt motion presets allow rapid changes to an actor's movement, such as faster running, with keyframes defining start and end positions for real movement.
• Props and a large library provide scene detail without rebuilding from scratch, letting users add street lights and other elements to enrich blocking.
• Rendering options include snapshot versus render, with the latter offering frame specific exports and higher control over final visuals.
• Revisions are supported by the canvas agent, enabling iterative improvements without starting prompts from scratch.

Questions & Answers

Q: A real search query question about the video (How to, What is, Why does). Question 1

How can I use 3D Builder to plan and block a scene before generating AI video? The video shows starting with a rough blocking in the canvas, assigning a character, setting motion, and distributing props to outline the shot list. You can add keyframes to move a character from point A to B, then adjust timing and easing. This reduces the number of failed attempts and credits spent on unproductive generations.

Q: Question 2

What is the role of previs in AI video according to the video? Previs is described as a cheap, reusable studio approach that lets you stage shots with a fixed set, camera angles, and timing before any AI rendering occurs. This ensures consistency across shots and minimizes the variation that comes from requesting many shots individually.

Q: Question 3

How does the system handle labeling and identifying scene elements? The tool auto labels all elements present in the scene when you describe details at the beginning, and you can toggle these labels on and off to carry over to exports. This labeling helps you revise elements accurately without re guessing which object is which.

Q: Question 4

What options exist for cameras and movement in the 3D Builder workflow? The video explains two paths: use camera presets to instantly apply an angle or move the current view camera to user defined positions, and have the camera follow a target or object with a look at option. You can also adjust easing during camera moves.

Q: Question 5

How do you export and iterate on the AI video in this workflow? You export as either a snapshot or a render, choosing a frame for the snapshot or a full video for render. Then you place the renders in a canvas, wire references, and make revisions by describing changes to the agent which can update the scene efficiently.

Q: Question 6

What is the cost structure for using 3D Builder according to the video? Building a scene without prompting the agent to generate costs zero credits, while generating the scene with the agent consumes credits. The scenario emphasizes paying for the final rendered outputs rather than every attempt at generation.

Q: Question 7

How does this workflow apply to non film content like ads or episodic content? The video suggests that this planning method is beneficial for product ads where the same set must stay visually consistent, or episodic content where a fixed room appears repeatedly. The idea is to keep location and lighting consistent across multiple shots.

Q: Question 8

What is the suggested plan to maximize use of 3D Builder and credits? The creator recommends a plan offering credits and features that enable many configurations, including a long term annual option with credits and access to various model families, to support ongoing AI video projects while controlling cost and maintaining quality.

Summary & Key Takeaways

• The video explains using 3D Builder inside Topview Film Studio to previsualize AI video scenes before generation, reducing wasted credits and improving camera control, motion, and scene consistency. It highlights the benefit of staging shots first, then rendering, to ensure alignment with the director’s vision.

• The presenter demonstrates how to block out a scene with 3D objects, set up camera angles and movements, and apply motion presets, all while keeping the scene identical across multiple shots for coherent edits.

• The guide covers pricing notes, credits, and the zero credit cost if the scene is built without asking the AI agent to generate it, emphasizing a more affordable previs workflow for AI video projects.