LTX hits #3 Image-to-Video and #4 Text-to-Video on Artificial Analysis Video Arena (LTX announcement)
Lightricks’ LTX team announces their video model is ranked #3 on Image-to-Video and #4 on Text-to-Video on the Artificial Analysis Video Arena leaderboard. The tweet is purely a positioning/marketing post — two screenshots of the leaderboard and the tagline “Built for creativity. Designed for production.” No model variant, version, or technical detail is named. Relevant to the team as a competitive datapoint: LTX is a direct competitor whose open-weights line (LTX-2: Efficient Joint Audio-Visual Foundation Model) and product surface (LTX Studio Elements — tag, edit, and combine scene elements for visual consistency) are already tracked here.
Key claims
Section titled “Key claims”- LTX is ranked #3 on the Artificial Analysis Video Arena Image-to-Video leaderboard [tweet body, screenshot].
- LTX is ranked #4 on the Artificial Analysis Video Arena Text-to-Video leaderboard [tweet body, screenshot].
- The post positions LTX as among the world’s leading AI video models on Video Arena, framed as “built for creativity, designed for production” [tweet body].
Method
Section titled “Method”N/A — product announcement. The leaderboard itself (Artificial Analysis Video Arena) is a pairwise-preference Elo ranking over user prompts; no methodology change is being described here, just a placement update. Which LTX checkpoint is being scored is not stated in the tweet.
Results
Section titled “Results”Leaderboard placement only: I2V #3, T2V #4. No FVD / VBench / motion-consistency numbers, no head-to-head competitor breakdown, and the date the snapshot was taken is not specified beyond the tweet’s own posting time (2025-11-06). The 35.2K view count is product-marketing reach.
Why it’s interesting
Section titled “Why it’s interesting”A competitive snapshot for the team — LTX sits alongside Luma in the AI video space and is one of the few competitors that ships both open-weights research artifacts (LTX-2: Efficient Joint Audio-Visual Foundation Model — LTX-2 open joint A/V foundation model) and a consumer-facing product layer (LTX Studio Elements — tag, edit, and combine scene elements for visual consistency — LTX Studio Elements). A top-4 placement on a public preference-arena leaderboard is the kind of external evaluation that’s worth checking against Luma’s own internal evals, while bearing in mind the usual arena caveats (prompt distribution skew, judge-fatigue, no controllability/consistency axes). Pair with HumanScore: Benchmarking Human Motions in Generated Videos for an example of why arena-style preference alone misses motion-quality regressions.
See also
Section titled “See also”- LTX-2: Efficient Joint Audio-Visual Foundation Model — LTX-2 technical report; same vendor, open-weights joint audio-video foundation model.
- LTX Studio Elements — tag, edit, and combine scene elements for visual consistency — LTX Studio Elements product announcement from the same vendor, adjacent to Luma’s references-model work.
- HDR Video Generation via Latent Alignment with Logarithmic Encoding (LumiVid) — LumiVid HDR video generation work by the LTX/Lightricks team.
- HumanScore: Benchmarking Human Motions in Generated Videos — motion-quality benchmarking that complements preference-arena rankings.