> ## Documentation Index
> Fetch the complete documentation index at: https://docs.beeble.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# SwitchX 2.0
> Professional video-to-video AI generation with SwitchX 2.0
[SwitchX](https://www.youtube.com/embed/eZ5wedB7pV4)
> [!NOTE]
>
> SwitchX is available exclusively on **Beeble (Cloud app)**. It is not
> available in Beeble Studio.
> [!NOTE]
>
> This page covers **SwitchX 2.0**, selectable from the **Model** dropdown in
> the input panel. The default model is [SwitchX 3.0](https://docs.beeble.ai/beeble/switchx.md); 2.0
> remains available for longer shots (up to 480 frames) and its Camera
> Tracking On/Off control.
**SwitchX** is a professional video-to-video AI generation tool designed for filmmakers, VFX artists, and creators. Unlike standard video-to-video models that hallucinate or alter the entire frame, SwitchX uses your original footage's pixels as direct control signals.
The core principle of SwitchX is simple: **Switch anything, keep what matters.** We call it **Source-Locked Rendering**: your performance stays locked to your source footage, pixel for pixel and frame for frame, while only the light and the world around it change.
To get the best results from SwitchX, master two concepts: the **Alpha Mask** controls *what* gets changed, and the **Reference Image** controls *how* it looks.
***
## Source Footage
Source quality carries into the output, so feed SwitchX the highest-quality source you have. Native ProRes ingest is supported: upload your ProRes 422 (Proxy, LT, 422, HQ) `.mov` master directly, up to **4.5GB**.
***
## Alpha Masks
The alpha mask is how **Source-Locked Rendering** works: it tells SwitchX exactly what to lock to your source and what to generate.
* **White in the alpha (1):** kept from your original footage, frame for frame, and only relit and restyled to blend into the new environment.
* **Black in the alpha (0):** completely generated from your prompt and reference images.
SwitchX offers four masking modes:
| Mode | Description | Best For |
| - | - | - |
| **Auto** | Auto-detects and isolates the main subject. No manual input needed. | Background replacement, relighting, virtual production |
| **Select** | Manually select specific elements to keep or generate. | Targeted changes (e.g., new outfit, inpainting) |
| **Fill** | Selects the entire frame. Keeps geometry, alters aesthetics. | Full-frame relighting or restyling |
| **Upload** | Upload a custom alpha matte from external software. | Pixel-perfect control from Nuke, After Effects, etc. |
### Auto Mode
> [!NOTE]
>
> **Best for:** Standard background replacements, relighting a subject, and
> virtual production.
The engine automatically detects and isolates the main subject in your first frame, then propagates and tracks that mask throughout the rest of the video. No manual selection is required.
### Select Mode
> [!NOTE]
>
> **Best for:** Precision control, changing specific elements (e.g., changing
> clothes while keeping the face and hands intact), and advanced inpainting.
You manually choose exactly which parts of the frame to keep. The objects you select become white in the alpha and stay from your original footage; everything else is black and gets generated. To generate inside a selection instead, use **Invert** (below).
Powered by interactive AI masking tools (SAM3, MatAnyone), choose a frame and click to select specific objects on it. The Alpha Editor lets you fine-tune your selections:
> [!TIP]
>
> * **Multiple Selections:** Select multiple objects. For the cleanest results,
> add distinct parts (like a face and a hand) as separate objects rather than
> grouping them.
> * **Deselect:** Right-click to remove an unintended selection.
> * **Invert:** Invert the mask of a specific object. For example, to turn a
> hand into a robot arm, select the hand, click **Invert**, and the engine will
> only generate within that area.
### Fill Mode
> [!NOTE]
>
> **Best for:** Complete scene relighting and restyling.
This mode selects the entire frame. Use Fill when you want to keep the original scene's geometry and composition intact but alter the overall lighting, mood, or artistic style.
### Upload Mode
> [!NOTE]
>
> **Best for:** VFX professionals using external compositing software.
Upload a precise alpha matte created in software like Nuke or After Effects. SwitchX reads the black-and-white alpha video pixel-by-pixel for absolute precision.
***
## Camera Tracking
SwitchX 2.0 supports camera tracking natively: the engine tracks the camera motion of your source footage and applies it to the newly generated environment. Control it under **Advanced** settings:
* **On (Follow Source)**: the default. The generated environment follows your source's camera motion.
* **Off (Ignore Source)**: the generated environment ignores the source's camera motion. Use this when you want the new background to stay stable regardless of how the camera moves.
Every environment below is fully generated, yet follows the source camera exactly — all with **Camera Tracking On**.
### Truck (Lateral Tracking)
The camera travels sideways with the subjects — the generated street scrolls past with the source's speed and parallax.
Source
SwitchX 2.0
### Pan (Follow)
The camera swings to follow the bike through the corner — the generated night circuit stays locked to the move even at race pace.
Source
SwitchX 2.0
### Push-in (Dolly)
The camera pushes forward past the foreground shelf — the generated aquarium keeps the same approach and parallax.
Source
SwitchX 2.0
***
## Fast Motion
SwitchX 2.0 holds detail through fast movement — lip sync, dance, gesture — where video models tend to smear or drift off the performance.
### Lip Sync
Every syllable lands after the switch from greenscreen to jungle — wardrobe and world change, the mouth doesn't.
Source
SwitchX 2.0
### Dance (Full-Body Motion)
Subway-platform floorwork relit from flat fluorescents to amber and signal green — the source movement carries through unchanged.
Source
SwitchX 2.0
### Gesture (Fast Hands)
A crossover dribble at full speed, switched from a street court to an indoor gym — the ball and both players stay sharp.
Source
SwitchX 2.0
***
## Reference Images
What your alpha covers keeps its pixels from your source, frame for frame. Your reference image and prompt control everything else: how that footage is relit and restyled, and what SwitchX generates around it.
The best reference is a finished frame: your subject composited into the new world, relit the way you want. Upload any image, or make one in the app with **Create with AI**.
### Imperfect References Are Fine
Your reference image doesn't need to perfectly match your source footage. If a face or a detail changes in a generated reference, that's fine. SwitchX knows what your original pixels are and brings only the lighting and style from the reference, applying it on top of your actual footage.
Source
Reference Image
SwitchX Result
***
## Settings & Export
* **Resolution:** 720p, 1080p, or 2160p. Each tier is a **maximum cap** on the shortest side, and a smaller source can also be enlarged (those options are tagged **Upscale** in the picker).
* **Length:** up to 480 frames (20 seconds at 24fps).
* **Export:** 8-bit MP4 and 10-bit MOV.
* **Aspect ratio and frame rate:** always match your source.
> [!NOTE]
>
> 2160p generation and the 10-bit MOV download require a **Professional** or
> **Team** plan.
### Iterate then Finish
Explore faster for less in 720p, pick your favorite, then **Finish** it in 4K. More than an upscale: Finish re-renders your take from the original source footage, recovering source detail to match a native 4K generation.
***
## Model Versions
| Capability | SwitchX 3.0 | SwitchX 2.0 | SwitchX 1.0 |
| - | - | - | - |
| **Resolution** | 720p, 1080p, 2160p | 720p, 1080p, 2160p | 720p, 1080p |
| **Max Frames** | 360 frames (15s at 24fps) | 480 frames (20s at 24fps) | 240 frames (10s at 24fps) |
| **Output** | 8-bit MP4 + 10-bit MOV | 8-bit MP4 + 10-bit MOV | 8-bit MP4 |
| **Reference Images** | Up to 4 | 1 | 1 |
| **Camera Tracking** | Built into the model (always on) | On (Follow Source) / Off (Ignore Source) | Not supported |
| **Iterate then Finish** | Explore in 720p, finish your favorite in 4K | Explore in 720p, finish your favorite in 4K | Not supported |
> [!NOTE]
>
> 2160p generation and the 10-bit MOV download require a **Professional** or
> **Team** plan.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.