Kling 3.0 Turbo · Image to Video

Read the price next to the plain 3.0 node before you buy the word "Turbo": $0.112 per second at 720p and $0.14 at 1080p, against 3.0's $0.1008 with audio at 720p and $0.1344 at 1080p. Turbo is the more expensive tier, and it has fewer widgets. That's not a mistake in the pricing sheet, it's a different product: you're paying for render speed and a leaner node, not for savings. If faster turnaround genuinely matters - a client waiting, a live-ish pipeline, a hundred clips to get through - Turbo is the right call. If you're iterating at your own pace, the standard node gives you more control for less per second.

What's on it (not much, on purpose)

• image_url - the opening frame, IMAGE, JPEG or PNG. The person, outfit and setting hold from this picture; the clip keeps its aspect ratio.
• prompt - what happens after the frame: the action and the camera. With audio always on, describe the sounds too. Dialogue in quotes gets voiced.
• resolution - 720p (default) or 1080p.
• duration_seconds - whole seconds, 3 to 15, default 5.

That's the node. No negative prompt, no guidance scale, no shot-type choice, no closing-frame input. Everything the standard 3.0 node exposes has been stripped back - which, to be fair, is a legitimate design: fewer dials is fewer ways to get a bad clip.

Audio is generated on every clip and there's no switch. Same as 2.5 Turbo in that respect: a matching soundtrack comes with the picture whether you wanted it or not. So the honest comparison at 720p is Turbo at $0.112/s against standard 3.0 with audio on at $0.1008/s - Turbo costs 11% more and gives you less to configure. At 1080p: $0.14 against $0.1344.

Outputs: video (VIDEO, with the generated audio inside the file - Save Video or a preview node) and task id (STRING, the console handle for the run).

Install

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

Restart ComfyUI, open Settings > SpicyAPI and paste a key from spicyapi.ai/console/keys, then restart once more so the node list is fetched from your account's catalogue. Manager search is "SpicyAPI". On a headless or remote machine, set SPICY_API_KEY in the environment. The pack has no dependencies of its own and downloads no models - generation happens on SpicyAPI's servers, so this runs the same on a laptop as on a workstation.

Practical notes

Nothing here is a free retry. Set Max cost per run in the SpicyAPI settings before you start: every job is quoted first and aborts uncharged if the quote exceeds your ceiling, then the node prints the actual charge when it finishes. A 15-second 1080p clip is $2.10, and with no resolution-y widgets between you and that number it's easy to queue one by accident.

Long clips on a lean node is the awkward combination. Without last_image_url or a shot list, a ten- or fifteen-second Turbo render is the model's own idea of continuity for that whole duration, and identity drift over long takes is the general weakness of video models, not a Kling-specific one. Three to five seconds is where this node looks best.

Cancels, seeds and reruns. Interrupting stops your wait, not the server-side task: it runs to completion, is charged if successful, and its result sits in your SpicyAPI console for up to 14 days. Seeds default to fixed so an accidental re-queue reuses the first result rather than billing you again - switch control after generate to randomize for a new take.

If it won't run at all: No SpicyAPI key is set means the key didn't make it into Settings > SpicyAPI (or SPICY_API_KEY); 40310 means verify your email; 40004 means the option combination you picked isn't served - change the field the message names. A model that's missing from the node search needs a ComfyUI restart, since the list is rebuilt from your account at startup.