Kling Cost Estimator

Kling Cost Estimator tells you, before you hit run, roughly how many credits a Kling generation will cost. This is the node that exists because the community's number-one objection to API-driven generation is "how much is this actually going to cost me?" - the same complaint that shows up in every thread about cloud video models in ComfyUI. The estimator gives you an answer in the graph, before the money moves.

Full honesty from the start: it's a local heuristic, not a quote. The node does no API call, reads no pricing endpoint, and takes no auth. It's a small multiplication table coded from published Kling pricing "as of 2026-05," and the info output literally says "Heuristic - verify in dashboard." Think of it as a cost-consciousness widget - a way to stop yourself from habitually clicking "run" on a 15-second pro-mode video without a second thought - not as an invoice.

How it works

Five inputs, all simple:

• kind - what you're pricing: text2video, image2video, image, tts, upscale_image, upscale_video.
• model_name - free text, default kling-v3; the estimator recognizes the v2/v3 family (and image models) and falls back to a default for unknown names.
• duration_sec - 1 to 60, default 5. Drives video cost linearly (a 10s video ≈ 2× a 5s one).
• mode - pro or std. Pro is modeled as a flat 2× multiplier on std, which is a real Kling behavior, if an approximate one.
• n - number of images, only used when kind=image (1–9).

Outputs: estimated_credits (INT, the raw number) and info (STRING, a human-readable breakdown: ~70 credits (model=kling-v3, mode=pro, dur=10s, n=1). Heuristic - verify in dashboard.).

The underlying table is worth knowing so the number doesn't feel like magic: text-to-video base costs roughly 35 credits (v3) per 5s std, v2.6 is ~30, v2-master ~50, v1.6 ~20; images are ~4 credits each on v3; upscaling is 6 (image) or ~15 per 5s (video). TTS is ~1 credit per 5 seconds of speech.

Where it's genuinely useful

• Batch pipelines. If your workflow loops multiple generations, wire the estimator in with your planned duration and let it print the per-run cost before execution. It's a sanity check on automation that would otherwise spend silently.
• Decision hygiene. Wondering whether to use std for the test pass? The estimator will show you the difference between pro and std in one read.
• Budgeting a project. Run it once for each scene length, add it up, and you have a rough project estimate before you render a frame.

Its limits

• Pricing drifts. The table is dated and Kling re-prices; the real number in your dashboard can differ. Don't plan a production budget on this alone.
• No credits for unknown model names beyond a fallback, and no handling for the fancier endpoints (effects, lip-sync, avatar) - those kinds just aren't in the dropdown.
• It won't stop you from spending. It's advisory. The pack doesn't gate generations on a budget - that's still on you, the dashboard, or your own discipline.

Install

Part of ComfyUI-Kling-Direct:

cd ComfyUI/custom_nodes
git clone https://github.com/IxMxAMAR/ComfyUI-Kling-Direct

or ComfyUI Manager → search "Kling Direct" → install → restart. No deps beyond what ComfyUI bundles. It's a couple hundred lines of arithmetic that might just save you from a nasty surprise on your first pro-mode batch.