fal.ai vs Replicate vs wiro.ai: AI video API prices

For the same AI video model, the API provider you pick can change the price per second by up to 105%. fal.ai lists Seedance 2.0 Mini at 720p for $0.155 per second, while wiro.ai lists $0.075. Across 37 model and resolution pairs that more than one provider sells, prices differ in 16 and match in 21. All prices are list prices from each provider's own page, checked October 29, 2026. The cheapest 720p second we found in the whole list is $0.03 (Veo 3.1 Lite at fal.ai, audio off).

How many video models does each API sell?

Our price list covers 43 video models. The four aggregators (fal.ai, Replicate, wiro.ai, Sume) sell 40 of them between them. fal.ai lists the most at 33, then Sume at 29, wiro.ai at 21 and Replicate at 9. A maker's own API sells only that maker's models, so the counts are small: Vertex AI 3, Gemini API 1, Kling 3, MiniMax 5, BytePlus 3, Runway 2, Luma 1.

Who is cheapest per model at 720p and 1080p?

We took every model that at least two providers sell, at 720p and at 1080p, and found the lowest list price per second. In 21 of 37 comparisons every seller charges the same. The rest are below, largest gap first.

The clearest case is Seedance 2.0 at 720p, sold by 5 providers. BytePlus lists $0.15 per second. fal.ai lists $0.303, which is 2.0x as much, and Replicate lists $0.18. A 30 second video at those rates costs $4.50 at the cheapest and $9.10 at fal.ai. For the Seedance family in detail, see Seedance 2.5 vs Seedance 2.0.

fal.ai vs Replicate: which is cheaper?

fal.ai and Replicate both sell 9 model and resolution pairs in our list. fal.ai is cheaper in 4, Replicate is cheaper in 3, and they tie in 2. Neither is the cheaper provider as a rule, so compare the exact model. On Veo 3.1 Lite both list the same rate. On PixVerse V6, fal.ai lists about half of what Replicate does. On Seedance 2.0, Replicate is lower than fal.ai.

wiro.ai and Sume are the other two aggregators in the data. Across their comparisons Sume is cheapest or tied 25 of 32 times, wiro.ai 21 of 29, fal.ai 25 of 37 and Replicate 2 of 9.

Is the maker's own API cheaper than an aggregator?

Often it is not cheaper, but it is rarely more expensive. Where a maker or cloud and an aggregator sell the same model and resolution (15 comparisons), the official API is at or below the cheapest aggregator in 14 and strictly below it in 2. BytePlus ModelArk, which sells Seedance, is the main example: it lists $0.15 for Seedance 2.0 at 720p against $0.303 at fal.ai.

What the pricing pages say besides price

These are the non-price points we could confirm on each provider's own pages when we wrote this post (October 31, 2026). We left out anything we could not confirm, such as free credits and commercial terms, so check those pages before you commit.

• fal.ai: its documentation describes a queue API where you submit a request, check its status and fetch the result, plus webhooks. In our price list its video models are billed per second or per video, depending on the model.
• Replicate: its pricing page says you only pay for what you use, and its documentation has a prepaid credit billing page and webhooks. Many video models are billed per second of output.
• wiro.ai: task details return a list price total and the amount actually charged after any discount, so you can see both per request.
• Maker and cloud APIs: each has its own billing, quotas and sign-up. We did not confirm those terms for every one, so read the linked pricing page.

When should you use an aggregator and when the maker's API?

• Use an aggregator when you test several models. One key and one billing account cover many makers. Moving between models means changing a model name, not opening a new account.
• Use the maker's API when you commit to one model. At volume, check the official price first. In our list it is at or below the aggregator rate in most shared comparisons.
• Always compare the exact model and resolution. The same name can be priced differently at 720p and 1080p, and audio is often priced separately.
• Do not assume the biggest name is cheapest. The gap table shows the same model at very different rates between providers.

What is the cheapest AI video API?

By list price on October 29, 2026, the lowest rate we found is $0.03 per second at 720p and $0.05 per second at 1080p (Veo 3.1 Lite at fal.ai). A low rate is not the same as a good fit: these are different models with different output. We compare documented features and list prices, not how the video looks. The pricing tool lets you sort every row, and the Reel cost article shows what a 30 second video costs across models.

Raytake is a Mac app that makes short videos through a chat with an AI assistant. It shows each model's price before it runs, and it asks you to approve still frames before paying for video.