Beeble SwitchX: AI for VFX and Video Relighting (2026)

Beeble SwitchX is a video-to-video neural network for VFX: it replaces the background, changes the lighting and props in an already shot video while keeping the hero: their face, facial expressions, movement and sync with the sound. Unlike "text to video" generators, which draw a clip from scratch, SwitchX takes your source frame as a base and redraws only what you selected with a mask. According to the developer, Beeble, the model delivers a clip with new light, background, shadows and even a changed camera angle in about 5 minutes. In Pixyn, SwitchX works right in the browser, and new accounts get welcome tokens to start.

TL;DR

• What it is: a video-to-video (and image) neural network for VFX from Beeble. It changes the background, light and props in a finished video while keeping the hero.
• The main difference from generation: it is not "invent a video from text" but an edit of already shot footage. The source pixels work as a control signal, so the hero does not "blur".
• How it is controlled: two tools, an alpha mask (what to change, what to keep) and a reference image (what the new scene should look like: background, light, mood).
• Strong sides (according to Beeble): keeps identity, facial expressions and audio sync; brings the redrawn area and the kept part together with correct light; works fast; has an API.
• Who it is for: videographers, VFX artists, content creators, advertising and marketplaces: move the hero into a new environment without a film studio.

---

What Beeble SwitchX is and who is behind it

SwitchX is developed by Beeble, a team known for SwitchLight technology: AI video relighting and extraction of PBR passes (normal maps, albedo and so on), which lets you "relight" shot footage as if it were a 3D scene. Beeble positions SwitchLight 3.0 as a video model for generating PBR passes from any video, and the separate desktop app Beeble Studio runs it locally on a GPU for 4K rendering.

SwitchX is Beeble's next product and their first video-to-video model. If SwitchLight answered the question "how to relight what was shot", SwitchX goes further: it not only changes the light but also replaces the background, props and surroundings, essentially building a new scene around the kept hero. It was announced in February 2026; the model is available in Beeble's cloud app.

How video-to-video differs from generating a video from scratch

An ordinary "text to video" generator (for example, from the overview of AI video models) draws a whole clip from a text description. Every run is a new video, and you do not control exactly how the person in the frame will look.

SwitchX works differently. You give it a finished video and say what to replace in it. It does not redraw the frame anew: it uses the pixels of your footage as a direct control signal, analyzing the material frame by frame. That is why, according to Beeble, the hero does not "hallucinate": their identity, facial features and movement come from the original, and only the masked area (background, surroundings) is generated. This is a different class of tool for a different task: not "compose a video" but "reshoot the scene around my hero".

To animate a still photo or generate motion from scratch, see Kling: how to turn a photo into a video or Seedance for video.

---

How SwitchX works: mask and reference

All of SwitchX's controllability rests on two things. Understanding them means understanding whether the result will work or not.

The alpha mask: what to change, what to keep

The mask tells the model which part of the frame to redraw and which to leave untouched.

• The unmasked part (the foreground, usually the hero): the original pixels are kept but intelligently relit for the new scene, so the hero does not look "cut out and pasted in".
• The masked area is generated anew based on the reference and the prompt (a new background, surroundings, props).

According to Beeble, SwitchX automatically detects the main subject in the first frame and carries the mask through the whole clip, so manual selection is not required in a simple case. In Pixyn the mask has four modes: automatic, fill (relight the whole scene), select an object right in the frame, and uploading a ready alpha mask (for example, one made in Nuke or After Effects), which the model reads pixel by pixel for maximum precision.

The reference image: what the new scene should look like

The reference is a visual "blueprint" of the result: background, lighting, mood, props, costume. SwitchX extracts the style and light from it and applies them both to the new area and to the kept hero.

An important nuance from Beeble's official materials: even if the hero in the reference differs slightly from yours (for example, it was generated separately), SwitchX relies on the pixels of the source video and does not rewrite the identity: it takes the light and the setting from the reference, not the face. As Beeble recommends, for a serious result it is worth not just taking a reference but refining it: adjust the color grading, add props, set the mood in an editor like Photoshop, and only then feed it to SwitchX.

---

Strong sides of Beeble SwitchX

Keeps the hero: face, facial expressions, sound

The main reason to take SwitchX instead of an ordinary t2v generator. Most video-to-video tools "float": from frame to frame the hero loses resemblance to themselves. SwitchX uses the source footage as a pixel guide, so, according to Beeble, it keeps the identity, facial features and audio sync, moving the hero into a new environment without losing recognizability.

Control instead of "guessing"

Unlike generators that redraw the whole frame and often "make things up", SwitchX gives explicit control: the mask decides what to touch, the reference decides how it will look. This is closer to a VFX pipeline than to a generative lottery.

Fast

According to Beeble, the model delivers a clip with new light, background, shadows and a changed angle in about 5 minutes. For a VFX task where the alternative is a film studio or manual compositing, this is fast. In Pixyn you choose the output resolution: 720p or 1080p.

Correct light interaction

The kept hero is not just moved into a new background but relit for it: Beeble presents this as physically correct interaction of light with props, costumes and scene elements. This is what "cut and paste" compositing usually lacks.

There is an API

Beeble offers an API for SwitchX to embed it into your own pipeline: useful if video edits are needed in series rather than once.

---

How to use Beeble SwitchX in Pixyn

1. Sign up in 30 seconds via email, Telegram or Google. Create an account. New accounts get welcome tokens, so you can try it without paying.
2. Open the Studio, the main generation screen.
3. Choose the Beeble SwitchX model in the list of models.
4. Upload the source video (or a frame), the one where you need to replace the background or light.
5. Set what to change: the mask selects the area to replace, the rest is kept.
6. Add a reference and a prompt: describe the new scene, its background, lighting and mood.
7. Run the generation: the finished clip appears in the feed, and you can download it.

---

How to use it: examples and scenarios

Example 1: move the hero to a new location

You have a clip with a person, shot at home or on a green screen. The goal is to place them in a cinematic location for a music video or an ad.

1. Upload the video: SwitchX selects the hero with a mask itself.
2. Prepare a reference of the new scene (for example, a neon street at night) and refine its color and mood in an editor.
3. Describe the scene with a prompt and run it.

The hero keeps their face and movement, the background is replaced, and the light on the hero adjusts to the new setting.

Example 2: turn a room into a studio

You recorded a podcast or a talking head in an ordinary room, but you need the look of a professional studio.

1. Upload the frame or clip.
2. The reference is the desired studio interior with the right light.
3. The mask keeps the host, and the background and setting are redrawn into a studio setup.

Who it is for

• Videographers and VFX artists: background replacement, relighting and compositing without a studio and LED screens; a ready alpha mask from Nuke or After Effects can be uploaded directly.
• Content creators and bloggers: change the location or turn a room into a studio for a clip.
• Advertising and campaigns: move a model or a hero into the right environment, change props and mood for the brand.
• Marketplaces and product shots: change the surroundings and light around a product or a hero in a video card without a reshoot.

If there are many similar edits, it is convenient to automate them with a visual pipeline: set it up once and run it in one click. How this works is in the article What Workflow in Pixyn is.

---

What it costs in tokens

Beeble from 4 tokens per image and from 10 per clip. A video costs more than a single frame, so check the light on one frame first.

The studio shows the exact charge before the run: check against it rather than the article, because the catalog gets updated. Pack contents are on the pricing page.

FAQ

How is SwitchX different from a text-to-video generator?

A "text to video" generator draws a clip from scratch from a description. SwitchX edits an already shot video: it changes the background, light and props but keeps the hero, whose face, facial expressions and movement come from the original. It is a VFX tool for editing.

What are the mask and the reference in SwitchX?

The mask defines which part of the frame to redraw and which to keep (usually the hero is kept). The reference image sets what the new scene should look like: background, lighting, mood. According to Beeble, the result depends heavily on both: even the same footage with a different mask gives a different outcome.

How long a video can I process?

According to Beeble's specification, up to 240 frames per task. These are short clips; split longer footage into parts. Plans are on the pricing page.

Will the hero's face and voice be kept?

This is SwitchX's strong side: according to Beeble, the model keeps the identity, facial features and audio sync because it relies on the pixels of the source video rather than generating the hero anew.

Can I use the results commercially?

Yes: for advertising, music videos and content. Plans and terms are on the pricing page.

---

Ready to try Beeble SwitchX?

Sign-up takes 30 seconds via email, Telegram or Google, and new accounts get welcome tokens to start.

Create your first VFX clip in Beeble SwitchX →

Already signed up? Open the Studio, choose Beeble SwitchX, upload a video and set the new scene with a mask and a reference.

Sources

• Beeble, the developer of SwitchX