One produced size against four priced rungs
One card states what the released model produces, 1216x704, plus a rule that resolutions divide by thirty-two. One documentation page lists 360p, 540p, 720p and 1080p on a single quality parameter. A lattice of legal sizes against a ladder of four. As of 2026-10-25.
| What the pages state | LTX-Video | PixVerse |
|---|---|---|
| Published values | 1216x704, with a divisor rule | 360p, 540p, 720p, 1080p |
| Cheap draft available | No tier to drop to | Two rungs below delivery |
| Exact pixel counts | Printed, and constrained | Not printed; names only |
Inclusion rule. Resolution statements from each page in the form each uses. A tier name is not converted into pixels, and a pixel pair is not converted into a tier. Order. The published values first, then what they allow.
1A ladder is an iteration budget
From 360p to 1080p is three doublings of area. Composition, blocking and timing can be judged at the bottom and only the placements actually bought need the top, which is how a campaign of variants is affordable.
The single-value row offers none of that. Every request returns the same frame, so there is no cheap attempt and no expensive delivery, only one cost repeated, and the exploration budget has to come out of the length instead.
2Pixels printed and pixels named
The single-value row prints its pixels and constrains them: both dimensions divide by thirty-two, so a placement asking for an exact frame may not be able to have it. The ladder row prints names and never the pixels behind them.
So one row can be checked against a published pixel minimum and cannot be asked for an arbitrary frame, and the other can be asked for a tier and cannot be checked. These sheets record which style each page used because the reader needs to know which check is possible.
3What else each page ties to the size
The ladder row ties a length to its top rung on two older model versions: ten seconds cannot be used at 1080p there. The single-value row ties nothing, because it has nothing to tie.
Read with the frame rules on the same card, that row describes a fixed research artefact and the other describes a configurable product with a version string that is part of the specification. Both belong in the column and neither is a grade of the other.
4The rest of the column behind these two
| Generator | Resolution |
|---|---|
| Amazon Nova Reel | 1280x720 |
| CogVideoX | 720 x 480 only |
| Google Veo | 720p / 1080p / 4k |
| Grok Imagine | 480p / 720p / 1080p |
| Hedra | 540p / 720p / 1080p |
| HeyGen | 128 to 4,096 px each side |
| Kling AI | 720p / 1080p / 4K |
| LTX-Video | 1216x704 |
| MiniMax | 480P / 768P / 2K |
| Pika | 480p / 720p / 1080p |
| PixVerse | 360p / 540p / 720p / 1080p |
| Runway | Pixel pairs, not tiers |
| SceneMixer | 480P / 720P / 1080P |
| Vidu | 540p / 720p / 1080p |
| Wan 3.0 | 480P / 720P / 1080P |
Inclusion rule. Every generator this register carries, on the one column this pair disagrees about. Rows that state nothing keep their place and are marked hollow. Order. Alphabetical by generator name.
Why these two. Both rows publish a resolution and one of them cannot be drafted on at all. The other's bottom rung sits below every published placement floor, so it is a working tier rather than a delivery one, and neither fact is visible in a single size figure.
- What the released model produces30 FPS videos at a 1216x704 resolution
- Quality parameter360p, 540p, 720p and 1080p
5Sources
Read from the Lightricks LTX-Video model card and the PixVerse text-to-video docs, checked 2026-09-22 and 2026-10-25 respectively. The column is described on resolution, and each entry has its own cell page at LTX-Video and PixVerse, and the rule that keeps surfaces and generators apart is on method. Nearby pairs: Hedra and HeyGen on length, Nova Reel and Grok Imagine.