# ComfyUI Lumi Batcher: parameter sweeps without editing nodes by hand > ComfyUI Lumi Batcher is a ByteDance extension for ComfyUI that runs a workflow across many parameter combinations at once, then presents the results in a multi-dimensional grid for preview and export. It is GPL-3.0 licensed, needs Python 3.10 or newer, and installs as a custom node. **bytedance/comfyui-lumi-batcher** — ComfyUI Lumi Batcher is a batch processing extension plugin designed for ComfyUI, aiming to improve workflow debugging efficiency. Traditional debugging methods require adjusting parameters one by one, while this tool significantly enhances work efficiency through batch processing capabilities. - Repository: https://github.com/bytedance/comfyui-lumi-batcher - Stars: 652 · Forks: 47 - Language: TypeScript - License: GPL-3.0 - Published: 2026-10-21 · Updated: 2026-10-21 · Language: en - Canonical page: https://hysenlabs.com/projects/bytedance-comfyui-lumi-batcher ## The problem is the twenty round revision loop The plugin is described as a batch processing extension for ComfyUI aimed at improving workflow debugging efficiency, and the documentation frames the problem as three recurring frustrations. The first is model indecision, spending hours replacing models by hand with no way to pinpoint which style is best. The second is parameter tuning, manually adjusting dimensions, weights or sampling steps through twenty revisions and still not meeting the requirement. The third is chaotic asset management, where generating a hundred images with inconsistent naming means ten minutes of generating and two hours of finding them. The claim is that traditional debugging requires adjusting parameters one at a time, while batch processing raises the work efficiency. That is a narrow and very recognisable pain, and it is worth being clear that this is a debugging and exploration tool rather than a production pipeline component. ## Beyond the three axis plotting model The plugin is positioned as going beyond tools like the XYZ plot available in WebUI, and the difference is stated in three points. First, it goes beyond basic dimensions, allowing you to cross-test any workflow parameter freely, with no restriction to XYZ axes. Second, flexible parameter grouping lets you combine parameters creatively, with the documentation giving the example of product shots and prompts crossed with different base models. Third, visual result management, where outputs are browsed in multi-dimensional grids for previewing and batch exporting. The documentation also says the plugin accelerates text, image and video creation workflows, and that the interface is meant to lower the learning curve for people who already know ComfyUI. ## Installing as a custom node The local environment requires Python 3.10 or higher, and there are two installation methods. The first is to clone the project into the ComfyUI custom node directory with git, then restart ComfyUI: ```bash git clone https://github.com/bytedance/comfyui-lumi-batcher.git ``` The second method is to search for `comfyui-lumi-batcher` in ComfyUI-Manager and click install. After installation, a button appears in the ComfyUI interface, by default in the upper right corner, which opens the batch tool interface. The repository structure reflects the shape of a ComfyUI extension, with an `__init__.py` at the top level, a `prestartup_script.py`, a `lumi_batcher_service/` directory for the backend, a `frontend/` directory, a `frontend-setup/` directory, a `static/` directory, plus `tests/`, `risk/`, `pyproject.toml` and `requirements.txt`. The pyproject also carries a Comfy Registry section with a publisher identifier and display name, so the extension is distributed through the Comfy Registry as well as through git and the manager. ## What the Python dependencies are used for The dependency list tells you what the plugin does with your results. `pandas` is pinned exactly at 2.2.0 in the project metadata, and combined with `openpyxl` it points to tabular export of the batch results, which is the spreadsheet friendly form of the sweep. `Pillow` at 9.0.0 or newer is the image handling, `aiohttp` is the asynchronous HTTP work, and `aiofiles` is the asynchronous file writing that the output side needs. `filetype` at 1.2.0 is there to sniff file types, which is how a single batch can contain images, video and audio without the tool guessing from the extension. `natsort` provides natural sorting, which is what makes a grid of outputs numbered 1 to 100 order correctly instead of 1, 10, 100, 11. The requirements.txt file lists the same set with looser lower bounds, while the project metadata pins the newer floors, so an installed environment can end up slightly behind the declared constraints. The primary language of the repository is TypeScript, with the Python side being the service and the front end being the interface. ## A release history dominated by upstream compatibility The release notes are the most informative part of the documentation, and they read as a list of compatibility work. Version 1.0.0 was the initial release with basic batch processing and multi-modal support. 1.0.1 fixed file relative path and path compatibility issues in Windows scenarios, and 1.0.2 added bilingual Chinese and English versions with a language switch. 1.0.3 added task management with deletion, magic expressions for quick input on custom parameters, and download support in the single image preview. 1.0.4 optimised the detection logic so starting a batch task from a workflow no longer actually starts an inference, and added support for the cg-use-everywhere plugin. 1.0.5 added batch processing in API node scenarios, custom output directories, and support for the Image Save output node of the was-node-suite-comfyui plugin. From 1.0.6 the notes are mostly about tracking upstream ComfyUI, with 0.3.44 and the validate_prompt change named, 1.0.7 and 1.0.10 following the latest ComfyUI version, and 1.0.8 fixing a scroll issue in the official UI template selection popup plus drag-and-drop of a result image back into the canvas to restore the accurate parameters. 1.0.9 fixed the zoom behaviour when switching between cells containing images or video and added arrow key navigation. 1.0.11 added audio scene batch processing. ## Where this is the wrong choice The limits are structural and worth naming before you install it. The first is scope. A batch tool is a development aid, and the cost of a single image generation is exactly what it is without the plugin, so adopting it for production work adds a dependency and a layer of indirection for no benefit. The second is the coupling to ComfyUI itself. The release history is a ledger of upstream churn, and 1.0.0 to 1.0.11 is eleven releases inside a period where the underlying application was moving from 0.3.44 to 0.3.67, with several releases existing purely to keep the batch tool working. That is a fair trade for a plugin, but it means you should expect to update the plugin when you update ComfyUI, and you should be wary of running a version mismatch. The third is the licence. GPL-3.0 is a fine choice for a plugin distributed alongside ComfyUI, and it is not an issue for personal or internal use, but it is worth understanding that the dependency carries copyleft terms. The fourth is the ecosystem of third party plugins it has to interoperate with, which the release notes show is a running negotiation with cg-use-everywhere and was-node-suite-comfyui. ## Conclusion ComfyUI Lumi Batcher fits a ComfyUI user who is iterating on a workflow and needs to see many parameter combinations side by side, especially when the parameters do not fit the three axis shape that the WebUI plotting tool assumes, or when the sweep combines a model with a prompt. It is the wrong tool for a single production run, since generating one image needs none of this, and it is the wrong tool if you cannot accept a GPL-3.0 dependency in your ComfyUI installation. Before you install it, confirm your ComfyUI version is close to the ones the release notes mention, since the plugin has repeatedly had to adapt to upstream changes, and read the beginner guide linked from the README. The last push was on 2026-10-10, the package version is 1.0.11, and the project publishes no GitHub releases. ## FAQ ### How do I install ComfyUI Lumi Batcher? You need Python 3.10 or higher. Either clone the repository into ComfyUI's CustomNode directory with `git clone https://github.com/bytedance/comfyui-lumi-batcher.git` and restart ComfyUI, or search for `comfyui-lumi-batcher` in ComfyUI-Manager and click install. A button in the top right corner of the ComfyUI interface then opens the batch tool. ### How does ComfyUI Lumi Batcher differ from the XYZ plot tool? The documentation says it goes beyond the XYZ plot in WebUI in three ways: you can cross-test any workflow parameter freely with no restriction to XYZ axes, you can group parameters creatively such as combining product shots and prompts with different base models, and you browse the outputs in multi-dimensional grids for preview and batch export. ### Which ComfyUI version does the current release target? Version 1.0.10 added support for the latest ComfyUI version 0.3.67 changes to ensure the batch tool keeps working, and earlier releases tracked 0.3.44 and the validate_prompt change. The plugin has needed repeated updates to follow upstream ComfyUI releases. ### What does ComfyUI Lumi Batcher do with the results? Results are browsed in multi-dimensional grids for previewing and batch exporting, with the Python dependencies including pandas, openpyxl, Pillow, filetype and natsort behind the tabular export, image handling, file type detection and natural ordering of numbered outputs. Version 1.0.5 also added custom output directories and support for the Image Save output node of the was-node-suite-comfyui plugin. ## Sources - [bytedance/comfyui-lumi-batcher on GitHub](https://github.com/bytedance/comfyui-lumi-batcher) - [Issues](https://github.com/bytedance/comfyui-lumi-batcher/issues) - [License: GPL-3.0](https://github.com/bytedance/comfyui-lumi-batcher/blob/master/LICENSE) - [README](https://github.com/bytedance/comfyui-lumi-batcher/blob/master/README.md) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/bytedance-comfyui-lumi-batcher