WanVideo ImageClip Encode (Deprecated)

Straight to the point: this node is deprecated - its own display name says "(Deprecated)" right in it. It was the original one-stop image-to-video encode for the WanVideoWrapper: hand it a reference image, a CLIP vision model, and a VAE, and it did the whole job - CLIP-encoded the image, VAE-encoded it into latents, sized everything, and produced the WANVIDIMAGE_EMBEDS for the sampler. Convenient, but bundling all of that into one node made it inflexible, so the pack split the work into separate steps. If you're building a new Wan I2V graph today, use WanVideoClipVisionEncode (with a CLIP vision loader) instead. This page exists because the node still shows up in older workflows and tutorials, and you should know what it did and what to move to.

What it did

Because it combined CLIP vision and VAE encoding, it needed both models plus the target geometry in one place, and it exposed strength knobs for each channel - how much the CLIP embedding steered versus how much the VAE latent did. That all-in-one design is exactly why it was retired: the modern split lets you reuse a loaded CLIP vision model, swap encode steps, and wire in the newer identity/reference nodes cleanly. A monolith can't do that.

The inputs and outputs (for reference)

• clip_vision (CLIP_VISION), image (IMAGE), vae (WANVAE) - the three things it needed to do everything at once.
• generation_width / generation_height (defaults 832 x 480) and num_frames (default 81) - output geometry.
• noise_aug_strength (default 0) - noise added to the reference; a little can help motion, too much drifts from the image.
• latent_strength / clip_embed_strength (both default 1) - how strongly the VAE latent and the CLIP embedding each condition the result. The two-channel balance that its successors handle in separate nodes.
• adjust_resolution (default true) - auto-snap to a supported resolution.

Output is image_embeds (WANVIDIMAGE_EMBEDS) → the sampler.

How to install it

It ships with the pack, so there's nothing special to install - but there's no reason to install the pack for this node. If you already have WanVideoWrapper:

cd ComfyUI/custom_nodes
git clone https://github.com/kijai/ComfyUI-WanVideoWrapper
pip install -r ComfyUI-WanVideoWrapper/requirements.txt

(or ComfyUI Manager → search ComfyUI-WanVideoWrapper), then restart. You'd still need a CLIP vision model in ComfyUI/models/clip_vision and a Wan VAE - the same files the current nodes use.

Common issues & troubleshooting

It errors after a ComfyUI or wrapper update. Deprecated nodes are on borrowed time against a fast-moving pack. Don't debug it - rebuild that part of the graph with LoadWanVideoClipTextEncoder → WanVideoClipVisionEncode, which is the maintained path.

You inherited an old workflow that uses it. If it still runs, you're not obligated to rip it out this second. But treat it as living on borrowed time, and any time you're reworking the I2V section anyway, migrate to the split nodes.

Trying to wire the newer identity/reference nodes into it. That's a big reason to migrate. Stand-In, BindWeave, and SVI expect the current WANVIDIMAGE_EMBEDS flow built from the separate encode nodes; the monolithic deprecated encode doesn't slot into those chains cleanly.