AI for video generation
Short videos from text or an image, such as reels, TikTok and ads. The list follows the Arena text to video score, with the price from each company's official page alongside, plus an open weights alternative.
Suggested models
Order: one main model per company, from the highest to the lowest score on Arena text to video, a public leaderboard of blind votes between pairs of outputs. The official price is shown next to it, and models without an Arena score go to the table below. Source: LMArena on Hugging Face, latest version we checked.
1 Gemini Omni 1.2 Flash Google
Why it fits: Google's main model to generate video. According to the official page: Duration: 3 to 15 seconds; up to 30 seconds with extension · Native audio: yes · Vertical format: yes, 9:16 · Resolution: 360p to 4K (official page)
- Arena text to video: 1,540 points · Arena image to video: 1,503 points (source: LMArena on Hugging Face, latest version we checked)
- Price: US$ 0.07 per second (source: Google, official page; approximate value at 720p; actual billing is by video tokens)?what it is
- Estimated cost (see the How we calculate block): about US$ 1.05 (estimate)
- Weights: Undisclosed?
- Alternative: open weights: LTX-2.5 · cheaper: Grok Imagine Video 1.5 Lite, US$ 0.02 per second
2 FLUX 3 Video Black Forest Labs
Why it fits: Black Forest Labs's main model to generate video. According to the official page: Duration: 5 to 20 seconds · Native audio: yes, with synchronized audio · Vertical format: yes, 9:16 and 9:21 · Resolution: HD to 4K (official page)
- Arena text to video: 1,491 points · Arena image to video: 1,451 points (source: LMArena on Hugging Face, latest version we checked)
- Price: US$ 0.15 per second (source: Black Forest Labs, official page; HD)?
- Estimated cost (see the How we calculate block): about US$ 2.25 (estimate)
- Weights: Undisclosed?
- Alternative: open weights: LTX-2.5 · cheaper: Grok Imagine Video 1.5 Lite, US$ 0.02 per second
3 Grok Imagine Video 1.5 xAI
Why it fits: xAI's main model to generate video. According to the official page: Duration: 1 to 15 seconds · Native audio: yes, with speech in sync with the lips · Vertical format: yes, 9:16 · Resolution: 480p to native 1080p (official page)
- Arena text to video: 1,488 points · Arena image to video: 1,453 points (source: LMArena on Hugging Face, latest version we checked)
- Price: US$ 0.08 per second (source: xAI, official page; price from the pricing table; the price per second changes with resolution)?
- Estimated cost (see the How we calculate block): about US$ 1.20 (estimate)
- Weights: Undisclosed?
- Alternative: open weights: LTX-2.5 · cheaper: Grok Imagine Video 1.5 Lite, US$ 0.02 per second
4 Seedance 2.0 ByteDance
Why it fits: ByteDance's main model to generate video. According to the official page: Duration: 4 to 15 seconds · Native audio: yes; silent video is also an option · Vertical format: yes, 9:16 · Resolution: 480p to 4K (official page)
- Arena text to video: 1,483 points · Arena image to video: 1,480 points (source: LMArena on Hugging Face, latest version we checked)
- Price: US$ 0.15 per second (source: BytePlus (ByteDance), official page; 720p; official example of a 16:9 video of 5 seconds, actual billing is by tokens)?
- Estimated cost (see the How we calculate block): about US$ 2.25 (estimate)
- Weights: Undisclosed?
- Alternative: open weights: LTX-2.5 · cheaper: Grok Imagine Video 1.5 Lite, US$ 0.02 per second
5 MiniMax H3 (linha Hailuo) MiniMax
Why it fits: MiniMax's main model to generate video. According to the official page: Duration: 4 to 15 seconds · Native audio: not stated · Vertical format: not stated · Resolution: 768P or 2K · Also: accepts reference images, videos and audio (official page)
- Arena text to video: 1,456 points · Arena image to video: 1,493 points (source: LMArena on Hugging Face, latest version we checked)
- Price: US$ 0.08 per second (source: MiniMax, official page; 768P)?
- Estimated cost (see the How we calculate block): about US$ 1.20 (estimate)
- Weights: Open?
- Alternative: open weights: LTX-2.5 · cheaper: Grok Imagine Video 1.5 Lite, US$ 0.02 per second
How we calculate
Change the numbers and the cost of every model on the page is recalculated right away. Without JavaScript, the default values below apply.
With these numbers: 15 seconds of video (estimate).
Cost of each model = quantity × the price per second of video published by the company. When the company uses another unit, the conversion is shown next to the price. It is a reference calculation and does not change the order of the list.
Other paid models we checked
Other versions and companies, with the price from each company's official page. They are left out of the list above when another model from the same company is already among the first five, when they do not meet this page's criterion, or when the company does not publish a price per second of video.
| Model | Arena score | Price and source | Estimated cost |
|---|---|---|---|
| Veo 3.1 Google | Arena text to video: 1,363 points · Arena image to video: 1,399 points | price not published (source: Google Cloud, official page; the Vertex AI table shows US$ 0.40 per billing unit, at 720p or 1080p with audio, without stating the unit in text) | no estimate: the company does not publish a price per second of video |
| Runway Gen-4.5 Runway | Arena text to video: 1,223 points | US$ 0.12 per second (source: Runway, official page; 12 credits per second, at US$ 0.01 per credit) | about US$ 1.80 (estimate) |
| Runway Gen-4 Turbo Runway | Arena image to video: 1,054 points | US$ 0.05 per second (source: Runway, official page; 5 credits per second, at US$ 0.01 per credit) | about US$ 0.75 (estimate) |
| Grok Imagine Video 1.5 Lite xAI | no Arena score | US$ 0.02 per second (source: xAI, official page) | about US$ 0.30 (estimate) |
| MiniMax H3 Lite MiniMax | no Arena score | US$ 0.04 per second (source: MiniMax, official page; 768P) | about US$ 0.60 (estimate) |
| Amazon Nova Reel Amazon | no Arena score | US$ 0.08 per second (source: Amazon, official page; us-east-1 region, official AWS price list) | about US$ 1.20 (estimate) |
| Seedance 2.0 Mini ByteDance | no Arena score | US$ 0.08 per second (source: BytePlus (ByteDance), official page; 720p; official example, actual billing is by tokens) | about US$ 1.20 (estimate) |
| Kling 4.0 Flash Kling AI | no Arena score | US$ 0.094 per second (source: Kling AI, official page; 720p) | about US$ 1.41 (estimate) |
| Kling 4.0 Kling AI | no Arena score | US$ 0.157 per second (source: Kling AI, official page; 720p, without input video) | about US$ 2.35 (estimate) |
| Sora 2 OpenAI | Arena text to video: 1,342 points | unavailable (source: OpenAI, official page; OpenAI shut the model down and it is no longer available in the API) | no estimate: the company does not publish a price per second of video |
FAQ
What is the best AI model to generate video?
By this page's rule, the list starts with Gemini Omni 1.2 Flash, FLUX 3 Video, Grok Imagine Video 1.5, Seedance 2.0, MiniMax H3 (linha Hailuo). Order: one main model per company, from the highest to the lowest score on Arena text to video, a public leaderboard of blind votes between pairs of outputs. The official price is shown next to it, and models without an Arena score go to the table below.
How much does it cost to generate video with AI?
With the default values in the How we calculate block, the estimate ranges from US$ 1.05 to US$ 2.25 among the listed models with a published price. It is only an estimate and you can adjust the numbers on the page.
Is there an open weights option to generate video?
Yes: LTX-2.5.
Does this page use AI to recommend?
No. The recommendation comes from fixed rules over public data from Epoch AI, OpenRouter, Hugging Face and each company's official pricing page, in the latest version we checked. The typed text is only compared with keywords.
Terms used on this page
Use cases with their own page
Official pricing pages checked
- Google, latest version we checked.
- Google Cloud, latest version we checked.
- MiniMax, latest version we checked.
- Amazon, latest version we checked.
- Runway, latest version we checked.
- BytePlus (ByteDance), latest version we checked.
- Black Forest Labs, latest version we checked.
- Kling AI, latest version we checked.
- OpenAI, latest version we checked.
- xAI, latest version we checked.
Sources
- Hugging Face Hub, metadados públicos do Hub; licença de cada modelo no repositório, latest version we checked.
Dates come from the sources: OpenRouter listing or Hugging Face repo creation, which may differ from the official launch. Family and version were grouped by rule from the name.