Updream 3D Preview Platform Tutorial - Controlling AI Camera and Camera Movement from Scratch
Friends, when you're making AI short films, do you often feel like you're playing an expensive gacha game? You're thinking of the protagonist elegantly turning and taking a step, but the AI generates a clunky, teleporting-like animation; you want that...
Friends, let's all do it! AI Do you often feel like you're playing an expensive gacha game when watching short videos?
My mind was clearly focused on the protagonist gracefully turning around and taking another step.AI What was generated, however, was a bizarre and grotesque animation resembling teleportation;
I want that cinematic, panoramic, push-pull camera style.Prompt wordsI wrote for a long time,AI I'll still do things my own way.
Many experts would suggest using Blender for modeling or C4D for positioning.
These specialized software programs require a long learning curve, and most ordinary people rarely have the energy to delve into them.
Fortunately, I recently discovered a super useful alternative that requires zero modeling skills and no complex code.Prompt wordsIt can also precisely control the lens.
This is the 3D preview platform that just launched on Updream. Let's take a look at the effect:
We can use 3D simulation to establish the spatial relationships between characters, adjust camera movements, and rehearse the scenes in our minds in 3D space before... AI 1:1 rendering reproduction。
Today's article will teach you step-by-step how to "direct" a play, allowing you to... AI To achieve precise, cinematic-quality camerawork.
We open Updream, and in the infinite canvas, right-click to create a new preview stage.
Building a virtual film set
In the preview stage, click on the generation space in the upper right corner and add a scene reference image.It's best to choose a pure scene.Images without people in them produce more stable results.
If the space is complex, you can also upload multiple scene images from different angles to stabilize scene details.
Agent They can analyze reference images and build a white model of the scene.
The roads, trees on both sides, and buildings appearing in the image are all identified with great accuracy; this level of detail and precision is among the best in its class.。
Arrange for actors to enter
We add characters to the scene.
Click "Add Character" in the upper right corner of the 3D preview stage.Add as many white character models as needed for the scene.。
By selecting the white model of the character in the scene layer, we can freely adjust its height, body shape, and basic pose.
They can accurately capture any height, weight, or body type.。
There are several commonly used ways to adjust character movement, one of which is...move(Shortcut key G), select to...Adjust the spatial position of the characterOne is rotation (shortcut key R), after selecting...Adjust the character's side profile and rotation angle.。
When adjusting, it's recommended to select a single axis line. The axis line will turn yellow when selected. Adjusting a single axis line is more convenient and accurate.recommendOverall adjustments are difficult to make precisely in 3D scenes.
After selecting a character, we can add the character's movement track below.Double-click the track to create a new path segment. Adjust the path segment length to match the desired video length.。
We can control the character's dynamics by dragging the white model of the character, drawing a path directly, or using keyframes.
For example, the action of a character walking forward along a road can be achieved in three ways:
Set up a camera
Now that we've determined the characters' movements, how do we film them?
We'll add a camera position, where you can set it up like a professional photographer: select the camera model, prime or zoom lens, lens focal length, and aperture.
Generally we do AI Short dramas and comics are the two most commonly used:
- For both distant and panoramic views: a 24mm wide-angle focal length + f/5.6 aperture provides a more stable depth of field.。
- Medium shot and close-up: 50mm focal length + f/4 aperture, minimal distortion, close to human eye perception, and can slightly blur the background to highlight the subject.。
The camera's position and trajectory are set up and adjusted in the same way as the person's. During the adjustment process...We can view the camera's perspective at that time through the preview box in the upper right corner.。
After we adjust the view from the first camera position, we can...One-clickCreate a following perspectiveThe camera position will always follow the composition of the first camera position for shooting.
We can click play to view the recording effect of the clip in real time.
Once you're satisfied with the preview, click the record button in the upper right corner. The recorded video of the white model will then be ready.automaticAdded to the canvas.
From blank model to finished product
The process of generating videos is familiar to everyone.
We are on the canvas,Add the white model video as a reference.Prompt wordsFocus on describing the art style, lighting, and clothing details; do not add descriptions of camera movement.,let AI Just follow the camera movement in the white model.
Let's take a look at the final result:
This follow-the-view perspective isn't difficult; we usually rely on...Prompt wordsIt can be made.
The advantage of a 3D preview stage is that we can freely define the distance and angle of the camera, and it is very precise, resulting in a stable composition.。
We can also manually adjust the camera position and create the camera movement effects we want:
This type of surround shot is often used in movies to highlight a character's moment of glory. If only relying on...Prompt words，AI It is difficult to accurately complete a 360-degree circular motion.
But in the Updream 3D preview stage,We can manually adjust the camera track; we can manually control the zoom and camera speed.。
The ability to precisely fine-tune shooting effects is the soul of professional video creation.
It's okay if the character's movement is a little complicated:
This case was very challenging, as the characters moved irregularly within a complex space.
If you rely on blind draws, the Fairy and Crane characters will most likely just fly around randomly.
Using the 3D rehearsal platform, I planned out the complex flight path in advance.AI They only need to handle rendering; the spatial logic is tightly controlled by the white model.。
Ordinary people can now handle the complex scheduling of such fantastical, large-scale scenes with ease.
When producing comics or short dramas, multiple camera angles can be added.
Each camera position, camera setup, and camera movement method are configured individually.
Some are responsible for setting the environment in the distance, while others are responsible for capturing emotions in close-up shots.
The combination of multiple tracks and multiple camera positions can make the transition between shots more natural and reduce the sense of disjointedness.
The combination of multiple tracks and multiple camera positions can make the transition between shots more natural and reduce the feeling of jarring scenes.
We ordinary video creators have finally gone from being just blind box collectors...Prompt wordsThe writer has returned to being a creator with a camera.
We used to AI It brings random surprises, but also suffers from the pain of not being able to replicate them.
updream AI It has built a bridge between the traditional film and television industry and simplified the expensive and complex professional pre-visualization process into a few clicks and drags on the canvas.
Even if we haven't learned any modeling software or understand complex photography theories, we can still shoot well-structured long takes based on intuition.
This sense of control is currently... AI The video field is extremely scarce.
We no longer need to repeatedly draw cards and burn money for the perfect shot. First, we can figure out how to shoot, deduce the tone, and then hand it over. AI The logic of reproduction is... AI Create what it truly should be like.