Supported providers

Sora (OpenAI), Veo (Google), Runway Gen-3, Pika, Kling, plus open-weight Hunyuan Video and Mochi for batch jobs via our compute fund.

Pricing

Video is billed PerSecondVideo with resolution tiers (480p, 720p, 1080p, 4K) and a quality multiplier. Cinematic-grade models cost dramatically more per second than fast-iteration models — Mode-level defaults select the right tier for your use case.

Job lifecycle

VideoGenJob → queued → running → succeeded / failed. Long-form jobs (>10 seconds) run async and emit a notification when complete. You can preview a low-resolution proxy while the high-resolution render finishes. Failed jobs are retried automatically once if the failure mode is transient.

Output storage

Generated video lands on Bunny CDN with optimizer applied. You get an HTTPS URL plus an embed snippet. Files are retained per your plan (Free: 30 days, Pro: 1 year, Business+: tenant-configurable).

Safety

NCII detection runs on every output regardless of tier. Realistic depiction of identifiable people requires either consent metadata on the generation request or a creator declaration that the subject is a public figure used under fair-use grounds.