Six tracks. One right choice.
waaave doesn't force every job through the same model. Stage 6 routes each content job through the visual generation track that fits the format, archetype, and product combination — automatically.
Start freeThe six visual generation tracks
LTX-Video 2.3 (default)
When: Default — when no other track is triggered
Product-focused prompt generated from your layout blueprint and Virlo trend brief. Fast inference, 9:16 output.
Runway gen4_turbo
When: When your product URL is available
Passes your product image as a visual reference to Runway's gen4_turbo model. Grounded in your actual product.
HeyGen Avatar
When: When you've selected 'Avatar' content archetype in onboarding
Generates a spokesperson avatar video using your chosen HeyGen avatar, lip-synced to the Stage 5 voice audio.
LTX-2.3 Fast (b-roll)
When: When the layout calls for ambient b-roll rather than product focus
Same LTX model as Track A but with a b-roll prompt — cinematic background footage rather than product closeups.
Slideshow (Reve 2.0 + Satori)
When: When the layout blueprint specifies 'slideshow' format
Generates per-section background images via Reve 2.0. Text overlays rendered as PNGs via Satori. FFmpeg composites into a final slideshow MP4.
Pinterest sourcing + Reve reproduction
When: When the layout calls for editorial-style imagery and no product URL is available
Playwright worker fetches source reference images from Pinterest. Each image passes through a Reve 2.0 reproduction pass — never hotlinked, always reproduced. IP-safe.
NSFW detection on every frame
Every generated image — from Reve 2.0 backgrounds, Pinterest-sourced frames, and thumbnails — passes through an NSFW detection filter before the job advances. Score threshold: 0.7. Jobs with flagged frames are blocked before Stage 7.