☁️BizyAir Wan_Model_Loader

This is the loader for BizyAir's separate text-to-video Wan path - distinct from the Diffusers pipeline nodes elsewhere in the pack, and it feeds a different downstream node (BizyAir Wan_T2V_Pipeline) with a different type entirely (WAN_MODEL, not the diffusers-style model_id string). Two different Wan integrations living in the same pack, worth keeping straight so you wire the right loader to the right pipeline.

What it loads

The checkpoint on offer here is Wan 2.1's 1.3B tier - the lightweight option, not the 14B flagship. Worth knowing what you're actually getting: 1.3B is meaningfully below 14B on quality, and it's the low-VRAM/drafting tier of the Wan family, the same weight class that also underpins things like Self-Forcing streaming generation elsewhere in the ecosystem. If you specifically want the 14B model, this isn't the node for it - BizyAir Wan_ImageToVideoPipeline loads Wan-AI/Wan2.1-I2V-14B-480P-Diffusers, the actual flagship, through the separate diffusers pipeline.

That's not necessarily a downgrade for your use case, though. 1.3B is genuinely useful for testing a prompt and a workflow shape cheaply before committing to a slower, pricier 14B run - and since this is all cloud-billed regardless of tier, running the small model first to validate your setup is a reasonable way to spend fewer credits before the real generation.

Inputs and outputs that matter

• ckpt_name - a single-choice dropdown, currently offering the 1.3B text-to-video checkpoint. There's no local file browsing here; you're selecting from whatever BizyAir hosts on its cloud.

Output is model (typed WAN_MODEL), which only makes sense wired into BizyAir Wan_T2V_Pipeline's matching input. This loader does nothing on its own - it exists purely to feed that one downstream node.

How to install it

• ComfyUI Manager - search "BizyAir", install, restart.
• Manual - cd ComfyUI/custom_nodes && git clone https://github.com/siliconflow/BizyAir.git, restart.
• Comfy CLI - comfy node install bizyair.

Set your API key before running anything - the in-app login prompt or an account at bizyair.siliconflow.cn. This loader executes against BizyAir's cloud regardless of the model tier, so it's billed the same as any other node in the pack even though 1.3B is the "cheap" checkpoint by local-hardware standards.

Common issues & troubleshooting

Output doesn't connect anywhere useful. WAN_MODEL is a BizyAir-specific type that only pairs with BizyAir Wan_T2V_Pipeline. It won't plug into the Diffusers-pipeline nodes elsewhere in this pack, and it won't plug into a generic Wan loader from another pack either.

Quality is noticeably below what you've seen from Wan elsewhere. That's expected from this specific tier - 1.3B is the lightweight/draft-quality checkpoint by design, not a bug in how BizyAir hosts it. If quality matters more than speed or cost for a given generation, this isn't the loader to reach for; the 14B diffusers pipeline is the better fit.

You expected a 14B option here and don't see one. Correct, there isn't one on this specific loader - the 14B model lives behind the separate Wan_ImageToVideoPipeline node using the diffusers-style model_id field instead of this WAN_MODEL type.

BizyAir turned up in your setup without a deliberate install. A known ComfyUI Manager bug used to mis-resolve missing nodes onto this pack, which is where most "is this safe" concerns about it online come from. It's a legitimate commercial pack from SiliconFlow - safe to remove through the Manager if you don't want it.