Wan 3.0 reads PDFs: what Sume's video request accepts

Wan 3.0 adds document and webpage inputs, covering PDFs, PPTs, Word documents and Excel sheets, according to its launch coverage. Sume's video request does not accept documents: it takes prompt, frame_images, input_references and related media fields. Summarize the document into a prompt yourself, then send it to wan-3.0.

What the launch post describes

The launch post also says Wan 3.0 recommends a video length from the prompt and includes an extension tool to lengthen videos. Those are model-side features as reported there. On Sume, you choose duration yourself, and wan-3.0 accepts 2 to 30 seconds.

What Sume's video request takes

The documented request fields are narrow, and this table lists them.

From a PDF to a clip

Wan 3.0 honors audio and video references on Sume, but none of these fields carries a PDF. Turn the document into text first. A short pipeline works well:

• Extract the text from the PDF with your own tool
• Summarize it into a scene description under a few sentences
• Pull one or two key images from the document and host them at public HTTPS URLs
• Send the summary as prompt and the images as input_references

Keep a fact check

Keep the document's facts in your own check step. A prompt that paraphrases a spec sheet can drift from the source, so review the output against the document before it goes out.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: doc-to-video-001" \
  -d '{"model":"wan-3.0","prompt":"Animated explainer of a three-step product setup, clean flat style","duration":12}'