Pipe In Context Video (WVW)

Pipe In Context Video (WVW) is the Wan-flavored context bus: same rgthree-style "everything on one wire" idea as the other context nodes, but built around the Wan Video Wrapper pack's types instead of plain ComfyUI ones. The "WVW" in the name is the giveaway - these nodes speak WANVIDEOMODEL, WANTEXTENCODER, WANVAE, WANVIDLORA, and the pre-computed WANVIDEOTEXTEMBEDS / WANVIDIMAGE_EMBEDS, which is exactly what ComfyUI-WanVideoWrapper produces and consumes.

Why does that distinction matter? Wan's 14B models are heavy, and the WanVideoWrapper flow deliberately separates the pieces so you can cache and reuse expensive parts - a WANTEXTENCODER and pre-computed text_embeds/image_embeds mean you encode your prompt and reference image once and reuse the embeddings across runs instead of re-encoding every time. This context node is the envelope that carries all of it: model, clip (the text encoder), vae, loras, text_embeds, image_embeds, plus the Wan loader knobs (frame_rate, frame_load_cap, skip_first_frames, select_every_nth, batch_size), images_input, a single images_ref reference image, images_output, audio in/out, two masks, the five images_pp_* post-processing slots and five any_* catch-alls, with steps, cfg, seed and path for settings. The single images_ref (rather than the start/end pair the generic Video nodes use) matches how the Wan wrapper handles reference conditioning.

Mechanically it's the same context dict as the rest of the family: optional base_ctx in, per-field overrides, unconnected fields inherit, and everything comes out both on the context wire and as individual outputs.

Install. ComfyUI Manager → search "RvTools" → install ComfyUI-RvTools_v2 and restart, or:

cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2
# restart ComfyUI afterwards

No models or keys ship with the node - it only carries Wan models, it doesn't download them. Manual clones: pip install -r requirements.txt (opencv-python, pynvml, piexif, Pillow, pilgram). You'll also need ComfyUI-WanVideoWrapper installed, or none of these WAN-typed inputs will have anything to plug into.

The catches. The big one: this node is useless without WanVideoWrapper - the WAN-typed inputs and outputs only connect to that pack's nodes. If you run Wan through ComfyUI's native support instead, this isn't your node; the generic Pipe In Context Video versions are. Second, it's the single-model WVW variant - the WVW v2 is the one with two models and the VRAM-management extras. And the standing caveats: the pack is unmaintained (ComfyUI_Eclipse is the README's successor), context is pipe-typed and needs a matching RvTools/rgthree-style Out node, and dict keys don't transfer across packs.