Nexus-BTA Transforms Into A Complete Local Creative Studio

    
        By vramkickedin    
     | 
    
            May 31, 2026 at 2:48 pm        
    
     | 
    
        2 min read

Nexus-BTA is a local AI studio that bundles image, video, workflow, and experimental 3D generation into one interface powered by an embedded ComfyUI runtime. The update tightens template handling, expands model compatibility, and adds an integrated visual node editor so users can tweak workflow graphs without leaving the app. It also deepens support for WAN 2.2, LTX 2.3 Director timelines, Qwen image editing references, and ControlNet presets across Flux and Z-Image.

Developer JpAndreBTA refined the UI after earlier feedback, focusing on smoother Civitai browsing, per‑model reference image slots, and a revamped inpaint canvas with generative outpaint. This release grows the tool from a generation front‑end into a more complete creative workspace while keeping the single‑click Windows launcher. The project remains primarily tested on NVIDIA/CUDA hardware, with Linux and macOS paths documented but not yet on par with the Windows experience.

Updated interface and expanded model support

Key improvements in this release

SD 1.5, SDXL, Flux, Qwen, and Z-Image pipelines.
LTX 2.3 linear view and Director timeline compositing.
WAN 2.2 generation with start/end frame control.
Visual node editor for inspecting and editing workflows.
Civitai browser with direct model download and sorting.
Inpaint canvas, outpaint, magic wand, and undo history.
Video upscale, frame interpolation, and background removal.
Experimental image-to-3D and texture‑paint workspace

This tool is built for prosumer GPU owners, serious hobbyists, and small agencies who run local AI on a single machine. Privacy‑conscious professionals can keep all generation offline while still accessing a managed ComfyUI backend with no cloud dependencies. By handling folder organization, template‑aware side menus, and output archiving automatically, it lowers the friction of switching between multiple model families and complex tasks.

What’s still experimental

The 3D workspace works for testing TRELLIS and texture‑paint routes but remains buggy and not fully optimized, with improvements planned for preprocessing and stability. Full LTX 2.3 local generation on macOS is not supported yet because the video pipeline still requires CUDA; Mac users can use SD/SDXL‑style image workflows but should treat LTX video tasks as NVIDIA‑only. The developer also notes that the Linux launcher scripts aren’t portable yet, so manual ComfyUI setup is required on that platform.