Pipeline

What happens after you press render

Five stages. Two of them exist purely to keep unlawful or unpublishable material from ever being made in the first place.

1. 01

   You write the prompt

   Describe the scene, the camera move, and the mood. Choose 5 or 10 seconds, an aspect ratio, a model tier, and whether you want generated audio. The credit cost is quoted before you press render.

2. 02

   Input moderation

   A deterministic rule filter runs first — it cannot be argued with by a cleverly worded prompt. An AI classifier then gives a second opinion. Blocked prompts stop here and cost you nothing, because no provider call is ever made.

3. 03

   Credits are held and the render starts

   Credits are deducted atomically so a double-click can never double-charge you. The prompt goes to Runway Gen-4.5 with your resolution, aspect ratio, and audio settings. Standard and Cinematic renders are silent; sound is only generated on the Cinematic + Audio tier.

4. 04

   Output review

   The finished clip gets a second safety pass. Clean output goes straight to your library. Anything ambiguous is held in a human review queue rather than released, and you are told it's under review.

5. 05

   Labeled export

   Downloads keep whatever watermark and provenance metadata the model provider embedded — we never re-encode to strip it — and every clip carries an AI-generated disclosure receipt you can attach when you upload.