OpenAI Videos compatibility
Call PopNow video models with the request and response shapes of the OpenAI Videos API.
If your code already speaks the OpenAI Videos API, you can point it at PopNow and keep most of it. The compatible endpoints live under https://api.popnow.ai/v1 and share keys, limits, validation and credits with the native API.
| OpenAI-style call | Does |
|---|---|
POST /v1/videos | Create a video task |
GET /v1/videos/{id} | Read its state |
GET /v1/videos/{id}/content | Redirect (302) to the finished file |
GET /v1/models | List callable models in OpenAI's list shape |
Only video, image and audio models are listed; chat and text endpoints are not part of the PopNow API.
Create a video
curl -X POST "https://api.popnow.ai/v1/videos" \
-H "Authorization: Bearer $POPNOW_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $(uuidgen)" \
-d '{
"model": "alibaba/wan-3.0/image-to-video",
"prompt": "Waves roll in slowly, the camera drifts toward the lighthouse",
"input_reference": "https://media.popnow.ai/movie/lighthouse.webp",
"seconds": "5",
"size": "1280x720"
}'How the fields map onto the model's own input:
| OpenAI field | Becomes |
|---|---|
model | A video model ID from /v1/models |
prompt | prompt |
input_reference | image_url / image_urls (an HTTPS URL or a popnow:// upload) |
seconds | duration_seconds |
size (WIDTHxHEIGHT) | resolution and/or aspect_ratio, or width / height when the model takes pixels |
A field the chosen model does not support is rejected with 400, never silently dropped. For settings OpenAI's shape has no field for, use the native createTask.
Read and download
States map as queued → queued, running → in_progress, succeeded → completed, and failed / expired → failed with an error object. GET /v1/videos/{id}/content answers 302 to a short-lived signed link — follow redirects in your client — 409 while the video is still being made, and 404 if it failed.
The returned objects never contain prices. Read the cost with the native recordInfo, using the same ID.
Errors
Errors use OpenAI's { "error": { "message", "type", "param", "code" } } shape with the same HTTP status the native API would return. A content refusal is HTTP 400 with code: "content_rejected", so you can tell it apart from a parameter error (invalid_request or unsupported_parameter, with param naming the field).