AI ranking
Video
Video generators (text-to-video and image-to-video) compared on output quality, prompt fidelity, control, cost, speed and generation artefacts.
Visual top 10
- 01 Heygen video 1 100.00
- 02 Gemini Omni 1.1 Flash 99.30
- 03 Bach 1.0 pro 98.67
- 04 Wan 3.0 96.98
- 05 Vimoo 1.0 95.55
- 06 Flux 3 Video 95.49
- 07 Dreamina Seedance 2.5 720p 95.48
- 08 Grok Imagine Video 1.5 Agent 95.29
- 09 Minimax h3 94.22
- 10 Gemini omni flash 93.29
Compare models. Tick up to 3 models in the table below, then run the comparison: who wins on which use, by how much, and how solid the measurement is.
| # | Model | Vendor | Score | Monthly subscription | Assessment | |
|---|---|---|---|---|---|---|
| 01 | Heygen video 1 | Heygen | 100.00 / 100 | Not available | Reliable | |
| 02 | 99.30 / 100 | €22/month | Reliable | |||
| 03 | Bach 1.0 pro | Video rebirth | 98.67 / 100 | Freemium | Reliable | |
| 04 | Alibaba | 96.98 / 100 | Freemium | Reliable | ||
| 05 | Vimoo 1.0 | Vimoo | 95.55 / 100 | Not available | Reliable | |
| 06 | Bfl | 95.49 / 100 | Not available | Reliable | ||
| 07 | ByteDance | 95.48 / 100 | Freemium | Reliable | ||
| 08 | xAI | 95.29 / 100 | Free | Reliable | ||
| 09 | Minimax | 94.22 / 100 | Freemium | Reliable | ||
| 10 | 93.29 / 100 | €22/month | Reliable | |||
| 11 | Minimax | 93.28 / 100 | Freemium | Reliable | ||
| 12 | Fal | 92.70 / 100 | Freemium | Reliable | ||
| 13 | xAI | 92.18 / 100 | Free | Reliable | ||
| 14 | Hidream o1 Video 1.0 | Hidream | 92.17 / 100 | Freemium | Reliable | |
| 15 | Bfl | 90.77 / 100 | Not available | Reliable | ||
| 16 | Bytedance seed | 90.29 / 100 | Freemium | Reliable | ||
| 17 | Bytedance seed | 88.61 / 100 | Freemium | Reliable | ||
| 18 | Meta | 88.22 / 100 | Not available | Reliable | ||
| 19 | Wan2.7 I2v | Wan | 86.56 / 100 | Freemium | Reliable | |
| 20 | Black forest labs | 84.81 / 100 | Not available | Reliable | ||
| 21 | Bytedance seed | 82.17 / 100 | Freemium | Reliable | ||
| 22 | Spacexai | 79.94 / 100 | Free | Reliable | ||
| 23 | Pixverse | 79.47 / 100 | Freemium | Reliable | ||
| 24 | NVIDIA | 79.25 / 100 | Not available | Reliable | ||
| 25 | Pixverse | 79.24 / 100 | Freemium | Reliable | ||
| 26 | Klingai | 78.36 / 100 | Freemium | Reliable | ||
| 27 | Klingai | 78.14 / 100 | Freemium | Reliable | ||
| 28 | xAI | 77.92 / 100 | Free | Reliable | ||
| 29 | 77.79 / 100 | €22/month | Reliable | |||
| 30 | Alibaba-ath | 77.75 / 100 | Freemium | Reliable | ||
| 31 | Klingai | 77.24 / 100 | Freemium | Reliable | ||
| 32 | Klingai | 77.02 / 100 | Freemium | Reliable | ||
| 33 | Alibaba-ath | 75.36 / 100 | Freemium | Reliable | ||
| 34 | 75.34 / 100 | Subscription | Reliable | |||
| 35 | Minimax | 75.01 / 100 | Freemium | Reliable | ||
| 36 | Bytedance seed | 75.00 / 100 | Freemium | Reliable | ||
| 37 | xAI | 74.89 / 100 | Free | Reliable | ||
| 38 | Pixverse | 74.84 / 100 | Freemium | Reliable | ||
| 39 | 74.34 / 100 | Subscription | Reliable | |||
| 40 | NVIDIA | 74.33 / 100 | Not available | Reliable | ||
| 41 | Agnes Video 2.5 | Sapiens ai | 73.99 / 100 | Not available | Reliable | |
| 42 | 73.83 / 100 | Subscription | Reliable | |||
| 43 | 72.14 / 100 | Subscription | Reliable | |||
| 44 | Kling | 71.70 / 100 | Freemium | Reliable | ||
| 45 | OpenAI | 70.93 / 100 | €23/month | Reliable | ||
| 46 | Bytedance seed | 70.10 / 100 | Freemium | Reliable | ||
| 47 | Klingai | 68.95 / 100 | Freemium | Reliable | ||
| 48 | Klingai | 66.65 / 100 | Freemium | Reliable | ||
| 49 | 66.53 / 100 | Subscription | Reliable | |||
| 50 | 66.25 / 100 | Subscription | Reliable |
Indicative score, weighted from public sources. Transparent methodology. See the AI ranking page for the full methodology and the other categories.
AI video generator ranking
Le Recul's Video ranking compares video generation models, in text-to-video as well as image-to-video. The grid is the same as for images — output quality, prompt fidelity, control, cost, speed, artefacts — but the trade-offs are far sharper: video adds temporal consistency, and that is where most models give way.
The demonstrations vendors publish are almost always short shots, selected from many attempts. In real use, the flaws appear over time: faces that drift, objects that deform between two frames, inconsistent physics, unpredictable camera moves. A model that impresses over five seconds can become unusable over twenty.
The score aggregates public signals weighted according to a « Video » grid, and every evaluation carries a reliability status depending on source coverage. The field moves very fast and its models are often in preview: read it as a snapshot, not as a stable league table.
What this ranking is for
This ranking helps you place a video generator before investing time or a subscription in it: the duration you can really use, the control you get over the result, the cost per generation and the rejection rate to expect.
It also helps put viral demonstrations in perspective: one successful shot says nothing about how many attempts it took, nor about how the model holds up on a longer or more constrained shot.
How to read the ranking
- Output quality and temporal consistency across the length of the shot, not on a single frame.
- Prompt fidelity: respecting the action described, not just the mood.
- Control: duration, framing, camera movement, consistency between shots.
- Cost and speed per generation, counting the failed attempts.
- How often artefacts, deformations and usage limits appear.
FAQ — Video
Which is the best AI video generator?
None dominates everywhere. The models rated highest on raw quality are not always the ones offering the most control or the best cost per successful shot. The ranking is there to help you weigh that by use.
Why is video harder to assess than images?
Because a video has to stay consistent over time. A model can produce excellent individual frames and fail on the continuity between them. That is the main cause of rejection in real use.
Are the announced durations realistic?
The maximum technical duration is rarely the duration you can really use. Consistency generally degrades well before the announced limit, which official demonstrations show little of.
Why are many models marked « Partial »?
Because public coverage of generative video remains thinner than for text. Many models are in preview, with few independent measurements available.
How often is the ranking updated?
About every 25 hours, automatically. The update date is shown at the top of the page.