Reliability, measured
Volume, success rate and generation speed — measured from our live job database, not marketing claims. Measured period: the last 30 days as of October 30, 2026.
All six models combined, upload to delivered MP4. Every one of them a finished, downloadable clip.
First-run success rate across all models. The other 1.9% was auto-refunded in full — see the failure breakdown below.
90th percentile: 5.7 min. Measured across all completed clips over the last 30 days, upload to delivery.
When a generation fails, its credit charge reverses automatically — no support ticket, no waiting.
Download promptly after generation; files are removed after a week.
One-time credit packs that never expire. Nothing renews, nothing to cancel.
Per-model breakdown
All six production models, last 30 days. Vidu Q4 and Veo 3.1 medians are higher by design — they render 16-second and 4K clips respectively.
| Model | Videos (30d) | Success | Median render | Known for |
|---|---|---|---|---|
| Seedance Mini | 2,318 | 98.1% | 4.0 min | Budget everyday clips, audio included |
| Kling 2.6 | 1,964 | 97.6% | 3.0 min | Cinematic motion, 720p |
| Grok Imagine 1.5 | 1,702 | 98.4% | 3.5 min | Fastest social clips, native audio |
| Vidu Q4 | 986 | 97.5% | 24.9 min | 16s long-form, up to 4K |
| Veo 3.1 Fast | 764 | 98.9% | 10.2 min | Maximum fidelity, 4K |
| Seedance 2.5 | 480 | 98.7% | 5.1 min | Native 30-second clips |
Why the other 1.9% fails
About 4 in 5 chargeable failures are prompts that trip upstream content filters — recognizable public figures, brand-protected characters, or flagged terms. Our pre-charge moderation screen catches most of these at submit time (blocked before any charge); the few that slip through to the upstream model are rejected there, counted as failures, and refunded instantly. The remaining share is upstream provider errors and timeouts — every one auto-refunded, never billed.
The practical takeaway: if a prompt avoids real-person likenesses and protected IP, its odds of first-run success are meaningfully higher than the 98.1% site average.
How we measure
- Render time is measured from job submission (after your image finishes uploading) to the moment the finished MP4 is delivered from our storage.
- A run counts as failed when the upstream model returns no usable video — content-filter rejections, provider errors, timeouts, or stuck jobs. Pre-charge moderation blocks happen before any charge, so they are neither billed nor counted.
- Success rate = delivered videos ÷ all chargeable runs (moderation blocks excluded from both sides).
- We re-publish these figures as the site grows. This page exists to be checkable.
Change log: first published October 27, 2026 · October 30, 2026: added 30-day volume, success rate, per-model breakdown and failure attribution.
The guarantee behind the numbers
AI video generation is probabilistic — no honest platform can promise every run succeeds. What we promise is the economics: you only pay for delivered videos. Failed runs reverse themselves, credits never expire, and there is no subscription to forget about.
15 free credits on sign-up · clips from 20 credits · no card required