Kling V3 vs PixVerse V6 vs Vidu Q3 (2026): API Price per Second, Tested

October 2026 brought three Chinese video-generation families online at NovAI in the same week: Kuaishou's Kling V3 lineup, PixVerse V6, and Vidu Q3 (Shengshu). Eleven new video models in total, all token-metered at the same upstream rate - which makes per-second cost directly comparable. This guide gives you the real numbers and the exact async workflow.

Price per second: the full October 2026 lineup

All eleven models bill tokens at $1.54 / 1M tokens (provider list rate, zero platform fee). What differs is token consumption per second of generated video, measured on real 5-second 768P generations:

How billing works

These models are token-metered, not per-request: you pay for the tokens the generation actually consumed, reported in the task result's usage.total_tokens. A failed task is refunded in full. Your balance is pre-checked against an estimate at submit time and settled against the real token count when the task succeeds.

The async task workflow (one endpoint for all eleven)

Submit a generation, poll the task, download the URL:

curl -X POST https://aiapi-pro.com/v1/video/generations \
  -H "Authorization: Bearer $NOVAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-video-v3",
    "prompt": "A red panda cycling through neon-lit Tokyo streets at night",
    "duration": 5,
    "resolution": "768P"
  }'

# poll until status = succeeded
curl https://aiapi-pro.com/v1/video/generations/TASK_ID \
  -H "Authorization: Bearer $NOVAI_API_KEY"

Swap kling-video-v3 for any of pixverse-video-v6.0, vidu-video-q3-pro, etc. - identical request shape across all eleven models.

Which family should you pick?

• Kling V3 / V3 Omni / O1 - cinematic motion and complex physics; Omni adds image-to-video conditioning; O1 is the reasoning-enhanced variant for multi-element prompts.
• PixVerse V6.0 / C1 - best cost efficiency in this group; C1 for drafts and high-volume pipelines, V6.0 for final renders.
• Vidu Q3 / Q3 Pro - strong temporal coherence; Q3 Pro when scene consistency across cuts matters.