Beeble Canvas gets a Topaz split

Canvas now compares Topaz upscales in-node, adds 30 plus generation models, Markdown notes, magnetic snap, and describer updates.
A man with a torch illuminates a rugged cave, casting flickering shadows on the textured rock walls. The warm glow of the flame contrasts with the cool, dark cavern surroundings, creating a mysterious ambiance as he explores the rugged terrain, clad in dark clothing.

For those who don’t know the tool: Beeble Canvas is a browser based node graph for AI VFX work across SwitchX, Image Generator, Video Matte, and Cloud tools. It connects to workflows around Nuke, Blender, and Unreal Engine.

The useful slider

Canvas now adds a built-in comparison view to its Topaz Upscaler node. Users can drag a split slider across the image and compare the original with the upscaled result directly inside the node.

A split-screen interface displaying the process of image upscaling. On the left, a low-resolution image features a woman with long brown hair, smiling during a conversation. On the right, an upscale tool with options appears, highlighting the model and settings, set against a sleek, dark layout.

That removes the extra Compare node from this specific check. Less graph furniture, fewer cables, same uncomfortable question: did the upscale add detail, or did it invent confidence?

For production artists, this is amazing: Upscaling lives and dies in small areas like eyelashes, hair strands, cloth weave, signage, texture noise, matte edges, pores, grass, compression mush, and the horrible little crawlies that only appear once the shot moves. A split slider makes those checks faster because the decision happens where the upscale happens.

Upscaling rarely fails politely. It can sharpen noise, clean away useful grain, hallucinate texture/moiré, or make AI material look more AI than befroe. It can also save a shot, rescue a low resolution element, or make a quick delivery possible without sending the plate through a separate app. The new split view does not guarantee a better upscale. It gives artists a faster way to judge one. That is a better promise.

A cleaner Canvas habit

An animated interface showcasing an AI image generation tool. The menu displays various model options, including 'Nano Banana 2' selected at the top. Below, an engaging preview image reveals vibrant colors and dynamic subjects, set against a dark background dotted with gridlines, adding a modern, tech-savvy aesthetic.

The update also adds more than 30 new generation models across video and image workflows. The visible Image Generator menu includes Nano Banana 2, Nano Banana Pro, GPT Image 2, GPT Image 2 High, Grok Imagine Quality, Grok Imagine, Seedream V5 Lite, Seedream V4.5, FLUX.2 Pro, FLUX.2 Max, FLUX.2 Flex, FLUX.2 Klein, and Recraft V4.1.

(deep breath….) The video side adds Grok Imagine, Grok Imagine 1.5, Wan 2.7, Wan 2.6, LTX 2.3 Pro, LTX 2.3 Fast, and Luma Ray 3.2. The listed modes include text to video, image to video, reference to video, video to video, Extend, Retake, and Reframe, depending on the model. That model buffet will get the screenshots. The Topaz slider will probably save more time. And waste less credits.

Notes, snap, describe

Sticky Note now supports rich text and Markdown. Artists can write headings, bold, italic, strikethrough, lists, blockquotes, inline code, and links. Pasted Markdown auto formats, and notes can be copied out as clean Markdown.

A square text box with a soft mint green background sits centered on a black dotted surface. The text is visually striking, showcasing formatting examples like bold, italics, and lists in a modern sans-serif font. On the left, a sleek vertical toolbar offers design options, enhancing the minimalist aesthetic.

That is proper graph hygiene. Notes often carry the part of the workflow nobody wants to reconstruct : prompt versions, reference choices, plate notes, supervisor picks, links to refernces “Somebody saw on a webpage”, failed settings, and why version 14 looked worse but passed review.

A man with a torch illuminates a rugged cave, casting flickering shadows on the textured rock walls. The warm glow of the flame contrasts with the cool, dark cavern surroundings, creating a mysterious ambiance as he explores the rugged terrain, clad in dark clothing.

The Compositor now has Magnetic Snap. Clips can snap to clip edges, the playhead, or frame 0. The magnet button toggles snapping, and the S key does the same. This is not glamorous, but neither is a frist frame slip in a client review.

Where this fits in the archive

This update plugs into the same Cloud direction covered in earlier Digital Production stories. Beeble Canvas introduced the node based AI compositor with SwitchX and SwitchLight under the hood. SwitchX API covered upload, generation jobs, polling, webhooks, 720p and 1080p output, alpha modes, prepaid USD credits, and pipeline integration.

Background Remover covered the standalone masking workflow with preview, refinement, RGBA export, and alpha only export. The Nuke plugin story covered PBR pass loading and relighting inside Nuke with Directional, Point, and HDRI lights. SwitchLight 2.0 covered PBR relighting, HDRI extraction, deflicker, neural detail enhancement, and local processing.

The pattern is simple. Cloud gathers model access, browser nodes, masking, generation, comparison, and timeline helpers. Local tools handle heavier pass based relighting and DCC handoff.



https://docs.beeble.ai/beeble/canvas

https://www.topazlabs.com/