Sora input image must match size: the Sume frame_images rule

Yes, in Sora's API the input image had to match the target video's size. On Sume there is no such pairing: you send the image as a first_frame entry in frame_images, and pick the output shape with resolution and aspect_ratio. A size field returns 400 unsupported_parameter on every v1 model.

Sora facts are from OpenAI's video generation guide; Sume facts are from Video generation, both read 2026-09-30.

What did Sora require for the input image?

The guide says an input image acts as the first frame of the video, is sent as input_reference in a multipart/form-data request, and "must match the target video's resolution (size)". Its example pairs a 1280x720 size with a 720p sample image.

How does Sume take a first frame?

frame_images holds "Images for first/last frames (image-to-video)", and each entry carries a frame_type of first_frame or last_frame. The separate input_references field is for style or content guidance. If both are sent, frame_images takes precedence and the request is image-to-video, so do not send both expecting a blend.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: first-frame-001" \
  -d '{
    "model": "seedance-2.5",
    "prompt": "Slow push-in on the product",
    "resolution": "720p",
    "aspect_ratio": "16:9",
    "frame_images": [
      { "type": "image_url",
        "image_url": { "url": "https://example.com/first.jpg" },
        "frame_type": "first_frame" }
    ]
  }'

Does Sume accept a size field?

No. The docs state that "every v1 model reports supported_sizes: null, so size returns 400 unsupported_parameter", and tell you to use resolution plus aspect_ratio. Each model advertises the values it accepts in supported_resolutions and supported_aspect_ratios.

What should I do with my image?

Pick an aspect ratio from the model's supported_aspect_ratios that fits your image, check it with GET /v1/videos/models before submitting, and crop the image to that ratio yourself if it differs. Sume's docs do not describe how a mismatched first frame is fitted, so a pre-cropped image is the predictable choice. For first and last frame on a specific model, see Seedance 2.5 first and last frame.