☁️BizyAir Veo3.1 Image To Video (Reference Images)

The plain Veo 3.1 image-to-video node animates one exact frame - whatever image you feed it is literally what the clip opens on. This one is different: you hand it up to three reference images and a prompt, and Veo uses those references to inform the generation rather than treating any single one as the opening frame. Think "generate a video featuring this subject/style" instead of "animate this picture."

That distinction matters for what you'd reach for each one for. If you need a specific still to be frame one of the output - a product shot, a specific pose - use the plain i2v node. If you're feeding Veo a character, an outfit, a location, or some combination of reference material and want it to figure out the composition itself, this is the one.

What it does

Same underlying model, same billing path as every other API node in this pack: this calls Google's Veo 3.1 through your BizyAir account, no local compute involved. The difference is entirely in how the image conditioning works - reference images instead of a literal start frame.

Inputs and outputs that matter

• prompt - the text description driving the video.
• ref_image_1 - required. Your first (and possibly only) reference image.
• ref_image_2, ref_image_3 - optional, if you want to combine multiple references.
• model - only veo3.1-fast is offered here. There's no pro option on the reference-image path, at least not yet - if you need the higher-quality tier, you'll have to work around it with the plain i2v node instead.
• aspect_ratio - only 16:9. No vertical option on this node, unlike the plain i2v version which offers both.

Those two restrictions are worth sitting with for a second: this node is genuinely narrower than its sibling. You're trading flexibility for the reference-conditioning capability. If you need 9:16 or the pro model, this specific node can't give it to you - check the plain I2V node before assuming a limitation is a bug.

Output is video plus bizyair_model_name, which currently just echoes "veo3.1" regardless - not useful for branching logic, just a log field.

How to install it

Identical to every other node in this pack:

• 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 doing anything else - click through the login prompt or register at bizyair.siliconflow.cn. Every request against this node spends credits.

Common issues & troubleshooting

You want pro quality or a vertical clip and the option isn't there. That's not a bug, it's the current schema - this node's model and aspect_ratio enums are genuinely smaller than the plain i2v node's. Switch to BizyAir_Veo_V3_1_I2V_API if you need either.

Results don't look like a clean blend of your references. Multi-reference conditioning in any video model is a "usually gets the gist, sometimes ignores one input" affair - there's no local knob to fix this since the actual blending happens inside Google's model. Try dropping to one or two references and see if consistency improves before assuming three is always better.

BizyAir shows up in your node list unprompted. Older versions of ComfyUI Manager occasionally auto-installed BizyAir when resolving a missing node from someone else's workflow, which is why you'll find "is this safe" threads about it online. It's a legitimate commercial pack, not malware - but if you didn't mean to add a cloud-billed dependency, it's safe to remove through the Manager.

Nothing runs at all. Check the API key first. Every node in this pack fails silently-ish without one - no local model to fall back on.