Anima Turbo LoRA · Image Editing

Same conversion job as the plain Anima Turbo edit node - photo or illustration in, anime out, strength controlling how much of the source survives - with one addition: you can attach up to three LoRA weight files, at their own strengths, to supply the look. About $0.018 an image, six-tenths of a cent more than the base node, which is a bargain if you already own Anima weights that do what you want.

This is the node for people whose "anime style" isn't a generic style. A specific artist, a specific show, a specific line-art treatment. Instead of fighting a prompt toward it, you attach the LoRA that already knows it.

Attaching the LoRAs

LoRAs come in through the loras input, and the source of them is a separate node: SpicyAPI LoRA. That node takes a direct download url - a .safetensors link that opens without a login, so a HuggingFace resolve URL or a public mirror, not a gated model page - and a scale from 0 to 4, default 1. Chain several SpicyAPI LoRA nodes together through their own loras input and the stack arrives at the model node in order.

One honest caveat about the Anima ecosystem: it moved fast. Preview 2 to preview 3 broke LoRAs badly enough that trainers had to redo them, and preview 3 to base 1.0 was contested - generic concepts survived, precise character and style detail didn't. If a LoRA that works locally produces mush here, that version mismatch is the first thing to suspect, not the scale.

The inputs

• prompt - describe the image you want back, not just the change. Tooltip from the author, and the right approach: when the LoRA supplies the style, the prompt should carry subject and composition.
• image_url - one source image. Aspect ratio is preserved from here.
• strength - 0 to 1, default 0.8. How far to move from the source. With a LoRA attached you can often go lower than you would otherwise, since the weights push style for you.
• resolution - 1k or 1.5k, where 1k is roughly a megapixel.
• seed - same seed and settings, same result; fixed by default so a re-queue doesn't charge you twice.
• loras (optional) - up to three Anima-trained weight files, each with its own strength.

Outputs: images and task id - the second being a STRING identifying the request for the console logs or a support ticket.

Install

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

Restart ComfyUI, then Settings > SpicyAPI and paste a key from spicyapi.ai/console/keys; on a server, set SPICY_API_KEY in the environment instead. Restart once more and the node list comes from your account. Manager users: search SpicyAPI. The pack installs no dependencies and downloads no models - the LoRA file itself isn't downloaded either, only its URL is passed on.

Cost, and a note on expectations

$0.018 per image, quoted before the run and printed at the end. strength gets you a whole range of looks from one source, and each attempt costs about two cents, so explore rather than over-think. Settings > SpicyAPI > Max cost per run remains your hard ceiling. Cancelling a run doesn't cancel the server-side task, and it is charged if it succeeds.

The failures you might see: No SpicyAPI key is set; 40310 for an unverified account email; 40004 if the model and option combination isn't served - the message names what to change and nothing is charged; and errors ending in request_id ..., which support will ask for. If the LoRA link is private, expect a rejection rather than a silent ignore: the URL has to open without a login.