NAI 图片转 Vibe 文件（旧版兼容）

The display name says it right in the title: NAI 图片转 Vibe 文件（旧版兼容） - "Image to Vibe File (legacy compatibility)". The source code marks it deprecated, and the README's Vibe flow has moved on to the cleaner two-node version: NAI 图片 Vibe 编码器 to encode, then a Vibe file saver to persist the file. This node is the old all-in-one that both encodes and writes a .naiv4vibe file in one step. It still works, it's kept around so older workflows don't break, but if you're building something new, reach for the encoder instead.

Mechanically it's the encoder with the file write bolted on: same core inputs - nai_pipe, image, information_extracted (default 0.7), strength (default 0.6), use_cache, paid_encode, force_reencode, optional mask - plus vibe_name for the filename. It does the paid encoding dance exactly like the encoder (you authorize via the button; a cache hit costs 0 Anlas), then always writes the .naiv4vibe file to the novelai vibes folder and appends the Vibe to the pipe.

Outputs: the updated nai_pipe, the Vibe文件名 string, and an 编码摘要 (cache hit vs. paid encode, filename, count of Vibes in the pipe).

The two differences that matter

• It always writes a file. The encoder has a save_vibe_file toggle and can just add the Vibe to the pipe in memory; this node's whole point is producing the .naiv4vibe artifact. If you only want the Vibe for the current run, this is extra work for nothing.
• V4 / V4.5 only. Because it writes a Vibe file in the V4 .naiv4vibe format, it refuses on V3 models and - same as everything Vibe - V5 isn't supported for Vibes at all yet. The error comes before any paid request, which is the pack's house style.

Install and verdict

Standard install: search ComfyUI-NovelAI-GENYTOOLS in ComfyUI Manager, or git clone into custom_nodes + pip install -r requirements.txt, token in settings.

My honest take: if a downloaded workflow has this node, leave it in place - it runs fine and the file it produces is exactly what the Vibe file loader expects. If you're starting fresh, the encoder + saver split is the better shape: encode once, cache it, and only save to disk when you actually want a shareable style file. This node is a compat shim, and it's a good sign the author keeps old workflows working - but you don't have to build new ones on legacy plumbing.