Ray2 vs Veo 3.0
Preferred for text to video: Veo 3.0 · Preferred for image to video: Veo 3.0 · Newest: Veo 3.0. Quality scores are blind human-preference votes; specs and prices come from each lab’s listings.
Where they differ
Quality is the arena score from blind, side-by-side human votes. Price, length and resolution come from each lab’s own listings.
- Preferred for text to videoVeo 3.0Ray2 1,065 (#46) · Veo 3.0 1,340 (#19) — blind human votes
- Preferred for image to videoVeo 3.0Ray2 1,108 (#46) · Veo 3.0 1,331 (#20) — blind human votes
- NewestVeo 3.0Ray2 Feb 20, 2025 · Veo 3.0 May 20, 2025
- Widest inputsSame inputsRay2: Text prompts, Reference images · Veo 3.0: Text prompts, Reference images
- Native audioNo native audioGenerates an audio track with the video: Ray2 no · Veo 3.0 no
- Image to videoBoth animate imagesStarts from an uploaded image: Ray2 yes · Veo 3.0 yes
- Self-hostingNo open weightsBoth are available only through APIs and apps
- Most API hostsVeo 3.0Ray2: 1 · Veo 3.0: 2
Every spec in one table
Dashes mean the data is not published.
| Specification |  Ray2Luma AI |  Veo 3.0Google |
|---|---|---|
| Quality (blind human votes) | ||
| Text to video arenaElo-style score · higher is better | 1,065 ±17#46 of 48 · 5,250 votes | 1,340 ±13#19 of 48 · 19,002 votes (best) |
| Image to video arenaElo-style score · higher is better | 1,108 ±16#46 of 48 · 9,521 votes | 1,331 ±12#20 of 48 · 32,367 votes (best) |
| Output | ||
| Produces | Video | Video |
| Inputs | ||
| Text prompts | Yes | Yes |
| Start image (image to video) | Yes | Yes |
| Video input | No | No |
| Audio input | No | No |
| Access | ||
| Weights | Proprietary | Proprietary |
| API hosts | 1 | 2 (best) |
| Released | Feb 20, 2025 | May 20, 2025 |
Which should you choose?
Which is better, Ray2 or Veo 3.0?
For text to video, voters prefer Veo 3.0 (Ray2 1,065 (#46) · Veo 3.0 1,340 (#19)). For image to video, voters prefer Veo 3.0 (Ray2 1,108 (#46) · Veo 3.0 1,331 (#20)). Arena scores come from blind, side-by-side human votes on the same prompts; when the gap is inside both models’ error margins we call it a tie. Test your own prompts before you commit.
What is the difference between Ray2 and Veo 3.0?
Ray2 is a video model from Luma AI (Feb 20, 2025), taking text prompts and reference images and producing video. Veo 3.0 is a video model from Google (May 20, 2025), taking text prompts and reference images and producing video.
Which is newer?
Ray2 Feb 20, 2025; Veo 3.0 May 20, 2025.
Which generates video with sound?
Generates an audio track with the video: Ray2 no; Veo 3.0 no.
Can I run any of these on my own hardware?
No open weights. Both are available only through APIs and apps.
