Get higher-quality clips from your videos with fewer, more complete highlights — powered by the next generation of Vizard’s clipping model.

What’s New

Vizard’s API has been updated to support clip_v2, the second generation of our AI clipping model. You can now specify which clipping model to use when submitting a project via the clipModel parameter — giving you direct control over the quality and behavior of the clips your pipeline produces.

The clipModel Parameter

When creating a project through the API, you can pass clipModel in the request body:

Note: If clipModel is omitted, clip_v1 is used by default. No changes to existing integrations are required.

Why Upgrade to clip_v2?

clip_v2 is built for teams that prioritize output quality over volume. Here’s what changes:

• Higher-quality clips — The model is optimized to surface more meaningful, well-structured highlights from your video.
• More complete clips — Each clip is more coherent and self-contained, reducing the need for manual review or trimming after the fact.
• Fewer total clips — Rather than generating a large number of candidates, clip_v2 returns a tighter, more curated set. This means less noise and less post-processing work for your team.

The tradeoff is a 1.25× billing rate compared to clip_v1. For workflows where quality matters — marketing content, highlight reels, social media publishing — this is typically worth it.

How to Use It

Add "clipModel": "clip_v2" to your existing project creation request. Here’s a full example:

curl -X POST https://elb-api.vizard.ai/hvizard-server-front/open-api/v1/project/create \
  -H "Content-Type: application/json" \
  -H "VIZARDAI_API_KEY: YOUR_API_KEY" \
  -d '{
    "lang": "en",
    "preferLength": [0],
    "videoUrl": "https://www.youtube.com/watch?v=OqLfw-TzzfI",
    "videoType": 2,
    "ratioOfClip": 1,
    "templateId": 52987165,
    "removeSilenceSwitch": 0,
    "maxClipNumber": 10,
    "keyword": "AI, spark, vizard",
    "subtitleSwitch": 1,
    "headlineSwitch": 1,
    "clipModel": "clip_v2",
    "projectName": "Introducing Spark 1.0 – Clip any video with a simple prompt now!"
  }'

The only change from a standard request is the addition of "clipModel": "clip_v2". Everything else remains the same.

Choosing the Right Model

Use clip_v1 when:

• You need maximum clip volume from a single video
• Cost efficiency is the priority
• You’re running high-frequency, batch-processing pipelines

Use clip_v2 when:

• Output quality is critical (e.g., client-facing content, social publishing)
• You want fewer, more polished clips with less manual curation
• You’re working with content where coherence and completeness matter

Billing

clip_v2 is billed at 1.25× the rate of clip_v1. The cost increase reflects the additional compute used to generate higher-fidelity clip selections. You can track usage per project through your existing dashboard.

Backward Compatibility

This update is fully backward-compatible. If your current integration doesn’t include the clipModel field, nothing changes — clip_v1 remains the default. You can adopt clip_v2 on a per-project basis at any time.

Summary

If you have questions about the new model or want help optimizing your integration, reach out to the Vizard developer team or check the API documentation.