Video Inputs

Not yet available. Video inputs are not yet available. This page describes the planned API.

Video input support will allow you to pass short video clips directly to multimodal models. The model processes the visual and audio track together and responds in text — useful for content moderation, video QA, scene description, and more.

Planned format#

Video inputs will use the same messages[].content array format as other multimodal inputs. A video_url part type will carry the clip:

{
  "role": "user",
  "content": [
    {
      "type": "video_url",
      "video_url": {
        "url": "https://example.com/clip.mp4"
      }
    },
    {
      "type": "text",
      "text": "Describe what happens in this video, second by second."
    }
  ]
}

Base64-encoded video will also be supported for private clips where a public URL is not available.

Supported formats and limits#

Planned supported formats: video/mp4, video/webm, video/mov. Maximum clip duration and file size will depend on the underlying model — check the model detail page for limits. Very long videos will likely need to be trimmed or chunked.

Supported models#

Video inputs will be available on models with the video capability. Query the catalog with ?capability=video once the feature is live, or filter by input_modality=video.

Use cases#

• Automated video captioning and description
• Content moderation pipelines
• Customer support screen recording analysis
• Educational video summarization
• News clip analysis in Arabic or French

Pricing#

Video inputs will be billed in DA per second of processed footage (rates vary by model). Check the pricing page and the model catalog for DA rates once the feature launches.

Related#

• Video Generation — generate video clips from text or image prompts
• Image Inputs — live today for still frames