☁️BizyAir Sora2 Image To Video

What it is

Sora is OpenAI's video model - the one that was, for a while, the benchmark every open video model got measured against before Wan closed most of the gap ("Wan 2.2 rated at the video arena higher than SORA," per one community comparison, though that's an older Sora/Wan matchup, not a claim about the specific Sora 2 build this node calls). This node wraps Sora 2's image-to-video API through BizyAir - you supply a starting image and a prompt, and OpenAI's servers generate a clip from there, no local GPU touched.

Like every node in this pack's "External APIs" category, this is a metered pass-through: you're paying OpenAI's rate through your BizyAir account, and you get zero say over what model architecture or weights are underneath it.

How it works

You give it an anchor image, describe what should happen in the clip, and pick between the standard and pro tiers. The job runs entirely on OpenAI's infrastructure; ComfyUI's role is just submitting the request and returning the finished video.

Inputs and outputs that matter

prompt describes the motion, action, or scene evolution you want. image is your required starting frame - the tooltip is direct that this is specifically the first frame, not a loose style reference, so composition and subject placement in this image carries directly into the output. model picks between sora-2 and sora-2-pro - the pro tier is presumably higher quality and higher cost, though the node itself doesn't spell out the difference beyond the name, so treat it as OpenAI's own tiering rather than something BizyAir defines.

The optional fields shape the output format: aspect_ratio (9:16 or 16:9, defaulting to 16:9), duration (10 or 15 seconds), and size (small or large) - a coarser control than the pixel-precise resolution options you'll see on some of this pack's other video nodes, worth knowing going in if you need an exact output dimension.

Output is video, your generated clip, plus bizyair_model_name, an internal tag identifying which underlying model handled the request - useful for logging, not for further wiring.

Installing it

Install via ComfyUI Manager (search "BizyAir") or clone manually: cd ComfyUI/custom_nodes && git clone https://github.com/siliconflow/BizyAir.git, then restart ComfyUI. Every node in this pack needs a BizyAir API key configured before it'll run - the login prompt on first launch handles registration.

Common issues

Sora's own content policy sits in front of this API, so expect rejections on anything it flags regardless of how the prompt is worded - that's OpenAI's filtering, not something BizyAir or this node controls. Because size only offers small/large rather than specific dimensions, don't expect pixel-perfect output control the way you might get from a node with an explicit resolution dropdown - plan your downstream compositing around whatever comes back. And as with every video-API node in this pack, if a generation fails or times out, check your BizyAir account balance first; these calls are billed per use, not free.