V1.1:

[Workflow]

• Updated the overall workflow.

• Qwen 2.1 is now used for watermark removal and high-fidelity upscaling. Prompt used: “Remove the watermark that [describe the watermark text or shape here]. Perform high-fidelity upscaling and fine denoising, removing compression artifacts and noise. Enhance the character linework, contours, hair strands, and clothing details, making the lines clean, sharp, and natural, while preserving the original anime art style, characters, composition, color palette, lighting, and shading exactly as they are. Avoid excessive sharpening, thickened lines, jagged edges, and any AI-generated redraw or re-rendering appearance.”

• Added BooruTagger + PixAI for prompt caption, significantly improving prompt accuracy.

• Plz put input image into input/ folder latest comfy changes only allow load image to load input folder

• 更新整体工作流。

• 使用 Qwen 2.1 进行水印移除与高保真上采样,大幅提高上采样精度和除噪效果。使用提示词：“去除[在此描述水印的开头或者形状]的水印.高保真上采样并精细除噪，去除压缩噪声与杂讯，增强人物线稿、轮廓、发丝和服装细节，使线条干净锐利自然，同时保持原有动漫画风、角色、构图、配色和光影不变，避免过度锐化、线条变粗、锯齿和AI重绘感。”

• 使用 BooruTagger + PixAI 进行提示词反推, 大幅提升提示词反推精度。

• 请将图片输入放入input文件夹，comfy更新后只能读取input文件夹文件。

V1.0:

This solution provides a ComfyUI‑based data captioning workflow that integrates SEEDVR upscaling, image bucketing, watermark removal, and WD14 auto‑captioning, complemented by a quick manual image cropping tool with keyboard shortcut support. Users can first crop images manually using the tool, then pass the cropped images into the workflow, which sequentially performs upscaling, bucketing, watermark removal, and WD14 captioning.
本方案提供一套基于 ComfyUI 的数据标注工作流，该工作流集成了 SEEDVR 图像放大、图像分桶（Bucketing）、水印去除以及 WD14 自动标注功能，并配套一款支持快捷键操作的图像快速手动裁剪工具。用户可先使用裁剪工具对图像进行手动截取，随后将裁剪后的图像送入工作流，依次完成放大、分桶、去水印及 WD14 标注.