Video Watermark (视频水印)

VideoWatermark is the burn-in node at the end of the pack's pipeline: take a generated clip, stamp text or a logo onto it, and get a video back. If you're posting Sora output to YouTube, Bilibili, or Douyin, this is how you make sure your handle rides along - and it does it locally with OpenCV and Pillow, no API call, no per-clip fee. That's a nice change of pace from the rest of ComfyUI_Sora: it's free, it's instant-ish, and it runs on whatever machine you already have.

How it works

The node reads your input video frame by frame, draws the watermark onto each frame using PIL/OpenCV (text with a TTF font, or your logo with optional background removal), re-encodes the result, and hands back the new file. Every parameter is applied per-frame, which is also how the animation effects work: fade, slide, rotate, scale, and blink are computed per-frame so the watermark moves instead of just sitting there. Local and deterministic - you can run the same clip through it a hundred times for free.

The inputs that matter

You don't need all 40+ knobs. The ones that do the work:

• watermark_type - text, image, or both. Everything else is conditional on this.
• text_content / font_size / font_color / text_opacity - what the text says, how big, which of the 9 colors (or custom with RGB), and how transparent.
• watermark_image - wire a Load Image in for a logo; image_scale (relative to video width) and image_opacity size and fade it.
• text_position / image_position / margin_x / margin_y - the 9-grid placement plus custom with custom_x/custom_y.
• animation - the fun one: 22 options from none through fade/move/rotate/scale/blink/slide, with animation_duration, move_direction, rotation_speed, and blink_frequency to tune it.
• start_time / end_time - show the watermark for only part of the clip; end_time = -1 means "to the end".
• add_shadow / add_outline - readability on busy footage. Keep shadow on (default true), it's cheap insurance.
• enable_multi_position - cycle the watermark between positions (four_corners, diagonal, all_nine, …) so a thief can't just crop one corner.

Outputs: video (VIDEO) and info (STRING).

Install

Standard pack install - ComfyUI Manager → search "ComfyUI_Sora", or:

cd ComfyUI/custom_nodes
git clone https://github.com/xuchenxu168/ComfyUI_Sora
cd ComfyUI_Sora && pip install -r requirements.txt

No API key, no config - just restart ComfyUI. opencv-python and Pillow come with the pack's requirements.

Common issues

• Watermark missing / video processing failed - make sure the input is a standard MP4 and the path resolves. The node needs ffmpeg available for re-encoding; if it can't find one, you'll see it in info.
• Chinese text renders as boxes - the README flags this directly: the default font may not cover CJK glyphs. Install a system Chinese font; the node uses your system fonts, so a missing CJK font produces tofu.
• Logo has a white box around it - turn on auto_remove_bg (with bg_color_mode white/black/auto and bg_threshold) to knock out a solid background before compositing.
• Animation looks janky - animation_duration too small makes moves jerky at low fps; give it 2–3 seconds and pick a single direction.

It's not glamorous, but it's the node that makes your output yours. For watermarking before you post, it's the one to reach for - and the advanced variant adds tracking if a fixed corner isn't enough.