Choosing a model
Both models turn an image plus a prompt into a video, share the same submit flow and per-second billing, and differ mainly in resolution, clip length, and audio.
At a glance
| Model | What it's best at |
|---|---|
| wan-2-2 | Wan 2.2 — high-quality open-source image-to-video. The economical choice for silent clips: product motion, social loops, ambient visuals. (up to 720p · clips to 15s · silent) |
| ltx-2-3 | LTX 2.3 — image-to-video with synchronized audio generated alongside the frames, delivered as one ready-to-post MP4. (up to 720p · clips to 20s · synchronized audio) |
| ltx-2-3-t2v | LTX 2.3 Text-to-Video (up to 720p · clips to 20s · synchronized audio) |
The submit form and the pricing tables always show the live durations, aspect ratios, and resolutions for each model — they come from the same configuration that prices your jobs, so what you see there is exactly what you can run.
When to pick Wan 2.2
- The clip will play muted (feeds autoplay silently, background loops, banners).
- You're generating in bulk and want the lowest per-second price.
- Short-form content where 480p/720p is plenty.
When to pick LTX-2.3
- You want sound without an editing step — ambience and audio matched to the scene.
- You need 1080p output.
- Longer storytelling clips.
Quality tiers apply to both
Every job takes a quality setting — Faster, Standard, High, Highest. Higher tiers spend more rendering steps on motion richness and detail, at a higher per-second price. Standard is the sweet spot for most work; use Faster for cheap drafts and High/Highest for final renders. Details in How pricing works.