Kling 2.5 Turbo · Text to Video

Native audio with video used to be the one capability with no open-source equivalent, and it's still the thing that makes hosted video models worth paying for. This node is the cheapest route into it in this pack: 1080p, 5 or 10 seconds, a soundtrack generated in the same pass as the picture, $0.056 per second. Five seconds of finished-looking, sounding clip for twenty-eight cents.

Text-to-video is the mode where that matters most, because you're describing a scene and hoping the model invents a coherent world for it - and the audio pass gives that invention a spine. Dialogue in quotes comes back voiced. Rain sounds like rain. It is not a mixing suite, and you'll still probably replace the track for anything with a real edit, but as a first pass it's remarkably hard to beat at this price.

What the node gives you

• prompt - scene, action, camera, and (because sound is always on) the sounds. Put spoken lines in quotes. The model handles a described ambience better than a list of adjectives.
• aspect_ratio - 16:9 (default), 9:16 or 1:1. Pick the frame before writing, since it changes how the composition gets planned.
• duration_seconds - 5 or 10, billed per second.
• negative_prompt - short list of what to keep out. Still does something on this generation, unlike the newer nodes where it's mostly decorative.
• guidance_scale - 0 to 1, default 0.5. Literalness, not CFG. Higher = closer to the prompt, lower = more latitude. If clips look stiff, this is why.

Outputs are video (a VIDEO with an audio stream - Save Video or preview; don't wire it to an image node) and task id (STRING, the run's handle in your SpicyAPI console).

Audio has no off switch. There's no generate_audio widget here, unlike the 2.6 and 3.0 nodes. If you want a silent 1080p clip, this isn't the node; if you want a cheaper sound-capable one, this is it.

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. Restart once more - the node list is built from your account's catalogue at startup, which is also why a model you saw on the site only shows up after a restart. Manager search: "SpicyAPI". For a cloud or server install, set SPICY_API_KEY in the environment and skip the settings panel. Dependencies: none beyond ComfyUI itself (Python 3.10+, the V3 node API).

Where this bites

No reference frame means composition roulette. Text-to-video can't hold a character, a framing or a specific object across takes; it can hold a place and a mood. If your shot hinges on a specific person or product, generate the still first and use the image-to-video node - you'll spend far less iterating at a few cents a still than at twenty-eight cents a clip.

Five seconds is the disciplined choice. Longer durations give the model more time to reinterpret what it started with, and the audio pass gets looser as it goes.

Cost discipline. Set Max cost per run in the settings before you start - each job is quoted first and anything over the ceiling aborts uncharged, and the node prints the real charge when it lands. Interrupting a run also doesn't abort the job: it keeps going on SpicyAPI, is charged if it succeeds, and its result sits in your console for up to 14 days. Seeds default to fixed, so re-queueing the same graph reuses the first result instead of billing you again; flip control after generate to randomize when you actually want a new sample.

The usual error codes. No SpicyAPI key is set (you skipped Settings), 40310 (verify your sign-up email), 40004 (that exact option combination isn't served - change the field the message names), and a proxy/firewall message if HTTPS_PROXY is set to something broken. Every error that ends with a request_id is worth keeping - hand it to support and they can find the request exactly.