Models

Each model you can call has a model ID that goes in the model field of createTask. The table at the bottom of this page lists every one of them with its starting price; each name links to a reference page with the full parameter list, price tiers and a request you can copy.

How a model ID is built

<maker>/<model>/<task>
alibaba/wan-3.0/image-to-video
google/nano-banana-2.1/edit
popnow/background-remover-v1/edit

• maker is who built the model. Tools built by PopNow itself use popnow.
• model keeps its version dots: seedance-2.0, not seedance-2-0.
• task says what the call does: text-to-image, edit, image-to-video, reference-to-video, video-extend, text-to-speech, text-to-audio and so on. Image editing is always edit.

A model family usually has several IDs, one per task. Pick the one that matches what you send: .../text-to-video takes only a prompt, .../image-to-video needs an opening image.

Read the catalog from code

The live catalog is the source of truth: it includes availability right now and the prices that apply to your account.

# Every video model, with input schemas
curl "https://api.popnow.ai/api/v1/models?modality=video&includeSchema=1" \
  -H "Authorization: Bearer $POPNOW_API_KEY"

# One model. The ID contains slashes; send it as it is.
curl "https://api.popnow.ai/api/v1/models/alibaba/wan-3.0/image-to-video" \
  -H "Authorization: Bearer $POPNOW_API_KEY"

Filters on the list: modality (image, video, audio), provider, task, search, plus includeSchema=1 and includeExamples=1 to add input schemas and validated example inputs. If your HTTP client treats a path parameter as one segment, encode the slashes as %2F.

Fields worth reading

Reading prices

credits is the same price popnow.ai shows, in credits (1 credit = $0.01):

"credits": {
  "unit": "per_second",
  "perUnit": 6.75,
  "rules": [
    { "when": { "*": "480p" }, "perUnit": 6.75 },
    { "when": { "*": "720p" }, "perUnit": 13.5 },
    { "when": { "*": "1080p" }, "perUnit": 27 }
  ]
}

unit says how the price scales (per_image, per_second, per_request, per_1k_characters), perUnit is the starting price, and each rule applies when the request matches its when values (* stands for the setting that decides the tier, here resolution). A task's final charge is rounded up to whole credits.

For the exact amount of one specific request, ask jobs/quote with the same body you will send to createTask.

Input schemas

inputSchema is standard JSON Schema (draft 2020-12). A few rules hold for every model:

• Unknown fields are rejected. A misspelt aspect_ration returns 400 instead of being silently ignored.
• Field names are shared across models. The same idea always has the same name: prompt, negative_prompt, image_url, image_urls, video_url, audio_url, duration_seconds, aspect_ratio, resolution, seed, generate_audio.
• Defaults are filled in for you. Send only what you want to change.
• Media fields take a public HTTPS URL, a popnow:// file you uploaded, or a saved asset. See Media.
• Display hints live in an x-ui object on each property (labels, order, slider steps). They never change validation; ignore keys you do not recognise.

If you leave out seed on a model that has one, a random seed is chosen and returned in the task's input, so you can repeat a result you like.

Caching

Catalog responses are private to your account and cacheable for 60 seconds. A sensible pattern is to cache the list for a few minutes, re-read a single model when its version changes, and treat available: false as "try again later or pick another model".

Every model you can call

Prices are the lowest tier per unit; higher resolutions, longer clips and extra options cost more. Each model's page shows every tier.