Veo 3.1 vs Kling 3.0: how to choose by inputs and limits

Choose by what the shot needs, not by name: Google's Veo 3.1 makes 4, 6 or 8 second clips with always-on audio and up to three reference images, while Sume's kling-3 entry makes 4 to 15 second clips at 720p or 1080p and takes no reference inputs. Sume lists kling-3 and does not list Veo, so a Veo-only need means Google's API.

Veo facts come from Google's Veo page, read 2026-09-29. Kling's own documentation site could not be read as text on that date, so the Kling column is Sume's catalog entry only, not Kling's full spec. This is a spec comparison, not a quality ranking.

How do their specs line up?

What does Sume list for each?

Sume lists kling-3 ("Kling Video v3 Pro") and no Veo model. The kling-3 entry takes text or first and last frames, duration 4 to 15, aspect_ratio 16:9, 9:16 or 1:1, and an optional generate_audio. It accepts no input_references. Check GET /v1/videos/models for the live list.

curl -X POST https://api.sume.com/v1/videos \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: kling-3-001" \
  -d '{
    "model": "kling-3",
    "prompt": "A skateboarder rolls past a mural at golden hour.",
    "duration": 6,
    "aspect_ratio": "9:16"
  }'

When should I pick which?

Run the same 5-second prompt on each and compare the frames you care about.

• You need to extend a clip you made: Veo 3.1 (Veo-made videos only).
• You need clips longer than 8 seconds in one call: kling-3 on Sume goes to 15 seconds.
• You need reference images to hold a subject: Veo 3.1 takes up to three; on Sume use a catalog model that lists image_url references, since kling-3 does not.
• You want one API for several models: Sume, where changing models is changing the model string.