Runway vs Kling vs Pika vs Luma API 2026 Compared

For a founder building video into a product, the four names that come up are Runway, Kling, Pika, and Luma. The headline question is cost per second, but the bigger gotcha is access: not all of them sell a self-serve API. This guide gives the verified per-second prices and explains how you actually get keys. All figures were checked against vendor and host pages on May 26, 2026.

1. Per-second API prices

Model	Per-second cost	How billed
Runway Gen-4 Turbo	$0.05/s[1]	5 credits/s @ $0.01/credit (direct API)
Runway Gen-4.5	$0.12/s[1]	12 credits/s @ $0.01/credit (direct API)
Kling 2.5 Turbo Pro	~$0.07/s[2]	$0.35 base 5s + $0.07/extra s (fal.ai)
Luma Ray 2	~$0.10/s[3]	$0.50 per 5s at 540p; 720p 2x, 1080p 4x (fal.ai)
Pika	Not publicly documented[4]	No self-serve API; partner or via fal.ai

Runway Gen-4 Turbo is the cheapest clean self-serve rate at $0.05 per second, with its higher-quality Gen-4.5 at $0.12 per second^[1]. Kling 2.5 Turbo Pro on fal.ai lands near $0.07 per second once you average the base clip and the per-second add-on^[2]. Luma Ray 2 on fal.ai is about $0.10 per second at base 540p, and rises with resolution^[3].

2. How you actually get API access

Pricing only matters if you can get a key. The four models differ sharply here:

• Runway: a clean self-serve developer API with credits at $0.01 each; sign up and call it^[1].
• Kling: the most practical route for most founders is fal.ai, which exposes Kling 2.5 with published per-second pricing; the official direct Kling API has historically required a large enterprise commitment^[2].
• Luma: a Dream Machine API billed separately from the consumer app, and also available through fal.ai with transparent per-second pricing^[3][5].
• Pika: consumer subscriptions only for self-serve; programmatic access is partner or via fal.ai^[4].

For a solo founder, the pattern is clear: Runway direct is easy, Kling and Luma are easiest through fal.ai, and Pika requires a conversation rather than a sign-up form.

3. The Pika and Kling caveats

Two of these names look like peers on a feature chart but are not on an access chart. Pika is a strong consumer video tool from around $8 per month, but it does not publish a self-serve API; you either reach out for partner access or call Pika models through a host like fal.ai^[4]. Kling makes excellent video, but its official direct API has been positioned for enterprise volume commitments, which is why most builders reach it through fal.ai's per-second pricing instead^[2].

The honest takeaway: if your roadmap depends on an API you control today, weight Runway and the fal.ai-hosted models, and treat Pika's direct API as not publicly documented until they publish self-serve terms.

4. Cost per finished clip

Per-second rates hide the real number, which is cost per finished clip after retries. A 5-second clip costs about $0.25 on Runway Gen-4 Turbo, $0.35 on Kling 2.5 Turbo Pro, and $0.50 on Luma Ray 2 at base resolution^[1][2][3]. But video generation often needs several attempts to get a usable take, so budget for a multiplier on these per-clip costs, not the single-generation figure.

Resolution compounds it: Luma Ray 2 doubles at 720p and quadruples at 1080p, so a 5-second 1080p clip is roughly $2.00 rather than $0.50^[3]. To model the monthly bill for your expected clip count and retry rate, use the AI stack cost calculator, and the ROI payback calculator to check whether the feature earns back its generation cost.

5. Which to pick

• Cheapest self-serve, easy access: Runway Gen-4 Turbo at $0.05/s.
• Higher quality, still direct: Runway Gen-4.5 at $0.12/s.
• Kling quality without enterprise contracts: Kling 2.5 Turbo Pro via fal.ai (~$0.07/s).
• Luma look, transparent pricing: Luma Ray 2 via fal.ai (~$0.10/s at base, more at higher resolution).
• Pika specifically: expect partner access or a fal.ai route; no public self-serve API.

Re-verify each price before you build; video model pricing and access terms change fast. For the budget-first angle, see the cheapest AI video API in 2026.

All pricing figures verified against official pricing and host pages as of 2026-05-26.

Frequently asked questions

What is the cheapest AI video API among Runway, Kling, Pika, and Luma?

Among models with straightforward self-serve API access, Runway Gen-4 Turbo is the cheapest headline rate at $0.05 per second of video, verified against Runway's API pricing in May 2026, with credits priced at $0.01 each. Kling 2.5 Turbo Pro on fal.ai works out to roughly $0.07 per second ($0.35 for a 5-second clip, then $0.07 per additional second), and Luma Ray 2 on fal.ai is about $0.10 per second at base resolution. Pika has no public self-serve API, so it is not a like-for-like option. The cheapest depends on the quality you need; Runway Gen-4 Turbo is the lowest per second, but higher-quality Runway Gen-4.5 is $0.12 per second.

Does Pika have a developer API?

Not a self-serve one. As of May 2026, Pika sells consumer subscriptions from around $8 per month, but programmatic access is through a partner or B2B arrangement rather than open sign-up, or indirectly via a third-party host such as fal.ai. If you need an API you can sign up for and start calling today, Runway, Kling (through fal.ai), and Luma are the practical choices, while Pika requires contacting their team for partner access. Build your stack around the providers that actually offer self-serve API keys.

How is video API pricing usually charged?

Most providers charge per second of generated video, sometimes via a credit system. Runway sells credits at $0.01 each and charges 5 credits per second for Gen-4 Turbo and 12 credits per second for Gen-4.5, so you multiply seconds by the model's credit rate. Aggregators like fal.ai charge a flat per-clip base plus a per-second add-on, for example $0.35 for a 5-second Kling clip plus $0.07 for each extra second. Resolution matters too: Luma Ray 2 lists a 540p base rate with 720p costing 2x and 1080p costing 4x, so a 1080p clip is materially more expensive than the headline number.

Should I call these models directly or through an aggregator like fal.ai?

For a solo founder, an aggregator is usually the faster path. Hosts such as fal.ai expose Kling, Luma, and other models behind one API key with published per-second prices, which avoids enterprise sales processes that some vendors require for direct access. The tradeoff is a small platform margin and dependence on the aggregator's uptime. Going direct can be cheaper at very high volume and gives you the official SLA. Runway is the main model here with a clean direct self-serve API, while Kling and Luma are often simplest to ship through fal.ai. Re-check both routes against your volume before committing.