Create video

Overview

Although this endpoint shares the videos path, we provide two distinct formats for differentiation:

Official Format - OpenAI's official API format

Sora-2 Custom Format Documentation - CometAPI's unique enhanced format

Supported Models

Currently supported model IDs:

sora-2-all - Supported video duration: 10 or 15 seconds (max 15 seconds)

sora-2-pro-all - Supported video duration: 10, 15, or 25 seconds (max 25 seconds)

Key Differences: Custom Format vs Official Format

Compared to the official format, CometAPI's custom endpoint has the following differences:

Duration Limits: Models with the -all suffix only support specific duration values for the seconds parameter:

sora-2-all: Only supports 10 or 15 seconds (25 seconds not supported) - Fixed rate of $0.08 per generation

sora-2-pro-all: 10, 15, or 25 seconds - Fixed rate of $0.8 per generation

Pricing Structure: Pricing differs from official format models

⚠️ Resolution Warning
When using sora-2-pro-all with seconds=25, there is a high chance that the requested 1024x1792 or 1792x1024 resolution cannot be output, and instead a lower resolution of 720x1280 or 1280x720 will be returned. To ensure high-resolution output, we recommend using 10 or 15 second durations.

Character Consistency Feature

Use the character_url and character_timestamps parameters to maintain character consistency in generated videos.

character_url string optional
URL link to a reference video for character extraction. The system will extract character features from this video and apply them to the newly generated video.

⚠️ The reference video must not contain real humans, otherwise the task will fail. We recommend using animated, cartoon, or virtual character videos.

character_timestamps string optional
Specifies the time range where the character appears in the reference video, format: {start},{end} (in seconds).

Example: "1,8" means extracting character features from second 1 to second 8 of the video.

Example Request:

📋 Storyboard Mode
The endpoint supports Storyboard functionality for creating continuous videos with multiple shots. When using storyboard, the prompt must strictly follow this format:

Shot 1:\nduration: 7.5sec\nScene: Airplane taking off\n\nShot 2:\nduration: 7.5sec\nScene: Airplane landing

Format Requirements:

Each shot starts with Shot N: (N is the shot number)

Use duration: Xsec to specify shot duration

Use Scene: to describe shot content

Separate shots with blank lines

Important Notes

⚠️ Asynchronous Processing
Video generation takes a long time, so it uses asynchronous mode. After creating a task, the system will immediately return a task ID. You need to use the query endpoint to get the generation progress and results.

⚠️ Content Policy
Generated video content must comply with OpenAI's usage policies. Generation of illegal, violent, pornographic, or copyright-infringing content is strictly prohibited.

🌟 Note:
Since this endpoint is implemented through technical means, please use sora-2-all or sora-2-pro-all model names, otherwise the actual call will be billed per second according to the official pricing.

curl --location --request POST 'https://api.cometapi.com/v1/videos' \ --header 'Authorization: ' \ --form 'prompt="cat"' \ --form 'model="sora-2-pro-all"' \ --form 'seconds="10"' \ --form 'size="720x1280"' \ --form 'character_url="https://filesystem.site/cdn/20251030/javYrU4etHVFDqg8by7mViTWHlMOZy.mp4"' \ --form 'character_timestamps="1,3"'

{ "id": "video_68e688d4950481918ec389280c2f78140fcb904657674466", "object": "video", "created_at": 1759938772, "status": "queued", "completed_at": null, "error": null, "expires_at": null, "model": "sora-2", "progress": 0, "remixed_from_video_id": null, "seconds": "4", "size": "720x1280" }

Overview#

Supported Models#

Key Differences: Custom Format vs Official Format#

Character Consistency Feature#

Important Notes#

Request

Responses

Overview

Supported Models

Key Differences: Custom Format vs Official Format

Character Consistency Feature

Important Notes