Model Selection

Model Comparison

Decision Tree

1. Need text-guided object targeting? Use pro.
2. Portrait/face-cam content? Use human.
3. Need max throughput and simple scene? Use light.
4. Otherwise default to original.

A/B Testing Recommendation

For new domains, test at least 30 representative clips across two candidate models before rollout.

Quality Guardrails

• Prefer representative source clips when tuning model defaults.
• Keep input compression moderate to preserve edge details.
• Use explicit output format policy for transparent vs non-transparent workflows.
• Run A/B validation with the same clip set before changing defaults.

Implementation Checklist

1. Define payload schema validation in backend before forwarding requests.
2. Store model/output_format/background settings with each job record.
3. Add internal quality review for difficult scenes (hair, glass, motion blur).
4. Create runbook entries for model-specific failure cases.

Model Rollout Plan

1. Define scene buckets (portrait, complex object, simple bg).
2. Benchmark all 4 models on each bucket.
3. Set fallback order by scenario and latency SLO.
4. Deploy with model override switch in ops console.

When to Use Model Selection

Model Selection belongs to the Guides section and covers map your use case to the best-performing model.

The page is written for developers and operators who need predictable video background removal behavior in production, not just a one-off demo request.

• Use this guide when model selection decisions affect output quality, processing cost, or user-facing workflow design.
• Test guidance on representative videos rather than only short demo clips.
• Record the chosen model, background, output format, and quality notes with each processed job for later debugging.

Implementation Notes

Before you promote this workflow, test it with at least one short clip, one longer clip, and one visually difficult clip from your actual product or customer segment.

For support and debugging, persist the original input reference, selected model, output format, credit usage, and final job status alongside your internal user or project id.

• Do not judge quality from a single frame; inspect motion, hair, hands, transparent edges, and fast turns.
• Do not compare model outputs using different source compression levels.
• Do not change production defaults without a rollback path and a known sample set.

Related Documentation

FAQ