Seedance 1.5 Pro Spicy · Image to Video

The "Spicy" suffix on any SpicyAPI node means the same base model with the content filter turned down. Here it's bytedance/seedance-1.5-pro-spicy/image-to-video - Seedance 1.5 Pro's image-to-video path, running looser. From $0.012 per second.

Whether that matters to you depends entirely on what you shoot. For a talking-head clip or a product turntable it's irrelevant; you'd be paying for a difference you'll never see. For anything the hosted model would parse and refuse - the kind of scene that keeps people running local video models in the first place - it's the whole reason this variant exists.

And here's the caveat that applies to every uncensored API model, including this one: there are no weights to patch, so the moderation you're routing around belongs to the endpoint. Whether the Spicy tier is a genuinely relaxed model or the same model with a different policy layer in front, you're trusting the provider's description. Test it before you build a project on it.

The one real difference from the standard node

The input list is nearly identical to Seedance 1.5 Pro · Image to Video, with one change that will bite you if you don't notice it: aspect_ratio defaults to adaptive, and adaptive is a real option here. The image-to-video sibling doesn't offer it. So this node will follow the ratio of the still you feed it unless you pick one of 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16 by hand.

Everything else is the familiar set. image_url is an IMAGE socket - wire Load Image into it. prompt describes the action, camera and mood. duration_seconds is 4 to 12 whole seconds, and it's the billed quantity. resolution is 480p, 720p or 1080p. generate_audio is on by default and produces a synced soundtrack; on this model it also moves the price. camera_fixed locks the framing when you want a stable shot. Optional last_image_url pins a closing frame, accepted together with the first frame only.

Two outputs: video (a normal ComfyUI VIDEO, straight into Save Video) and task id (a string, for the console or support).

seed arrives on fixed control-after-generate, so double-queuing a graph reuses the first result rather than buying a second one. That's a kindness in a paid node, not a limitation.

What happens on Run

The still is uploaded to SpicyAPI (PNG, with a JPEG fallback if it's over the 10 MiB image ceiling), the pack takes a price quote, submits the job and polls until the model is done. Then it downloads the MP4 and decodes it into the graph. Skipping the local GPU is the entire value proposition, and it's why this works on a laptop.

Nothing about it is offline. Prompts and images leave your machine by design, and the KB's read on this whole node category holds here: an API node is the right tool for a model you cannot run, and the wrong default for one you can.

Install

Search SpicyAPI in ComfyUI Manager and install, or:

cd ComfyUI/custom_nodes
git clone https://github.com/spicyapi-ai/comfyui-spicyapi

Restart, then put your key in Settings > SpicyAPI (or set SPICY_API_KEY on a headless box) and restart once more so the node list matches your account. No pip installs, no weights - the pack ships zero dependencies of its own.

Things that go wrong

The pack's own error table is short and honest. No SpicyAPI key is set means exactly that. 40310 is an unverified email - click the link from sign-up. 40004 means no deployment serves the option combination you picked (a closing frame at 1080p is the usual culprit on the 1.5 line); the message names the field to change. An error ending in request_id is the one to paste into a support ticket.

Two non-errors worth knowing: interrupting a run doesn't kill the job - it keeps generating server-side and is charged if it succeeds, with the result sitting in your console for up to 14 days. And a quote cap in Settings is the only thing standing between an experiment and a bill, so set one before you start queueing 12-second 1080p takes.