Runway API pricing in plain numbers: what one credit costs, how many seconds of Gen-4.5, Gen-4 Turbo and Veo 3.1 you get for $10, how to create API access, what each usage tier allows, and where free video generation actually exists. Sample budgets included.
Runway bills its API by the second of footage you generate, not by the month. You buy credits at $0.01 each, every model burns a fixed number of credits per second of output, and nothing renews on its own. That sounds simple until you open the pricing page and meet a dozen models, three resolutions and several audio add-ons. This article turns that page into a budget you can plan around: the per-second rates, what a five-second clip costs on each model, how to create API access, what each usage tier allows, and where free options truly exist (and where they do not).
💡 Quick answer: One credit costs $0.01. Gen-4 Turbo burns 5 credits per second ($0.05), Gen-4.5 burns 12 ($0.12), and Veo 3.1 with audio burns 40 ($0.40). The minimum purchase is $10, and the official docs list no free API balance.
Runway Dev, the developer platform, bills separately from the consumer web app. The pricing is published openly on Runway's developer docs, and it is refreshed often as new models arrive, so check it before you commit a large budget.
Credits cost $0.01 each, bought in the developer portal for a specific project, plus sales tax where it applies. The pricing page lists no seat fee and no platform subscription, only consumption. Every call to a model draws a fixed number of credits per second of video, or per image, and that rate is the only number you need to forecast spend.
A minimum payment of $10 is required to get started, which buys 1,000 credits. Two details trip people up:
How does that compare with a subscription? Third-party pricing roundups list the web plans at $15 a month for 625 credits (Standard), $35 for 2,250 (Pro) and $95 for 9,500 (Max). That works out to roughly $0.024, $0.016 and $0.010 per credit. The API's flat one cent matches the best web rate and beats the entry plans, although per-model credit costs in the web app can differ from the API table below. Confirm current plan prices on Runway's site before you rely on those figures.
💡 Treat the first $10 as a lab fee. It is enough to test three or four models properly before you commit to a bigger top-up.
Every video model has its own burn rate. The table converts the published credits per second into dollars and normalizes everything to a five-second clip so the models line up side by side.
| Model | Credits per second | Cost per second | Five-second clip |
|---|---|---|---|
| Gen-4 Turbo | 5 | $0.05 | $0.25 |
| Act-Two | 5 | $0.05 | $0.25 |
| Gen-4.5 | 12 | $0.12 | $0.60 |
| Wan 3 at 720p | 10 | $0.10 | $0.50 |
| Veo 3.1 Fast, no audio | 10 | $0.10 | $0.50 |
| Veo 3.1 Fast, with audio | 15 | $0.15 | $0.75 |
| Veo 3.1, no audio | 20 | $0.20 | $1.00 |
| Veo 3.1, with audio | 40 | $0.40 | $2.00 |
Some models only accept fixed durations, so read the five-second column as a comparison tool, not a promise of an available setting. Act-Two is the odd one out: it animates a character from a performance video rather than from a prompt alone, yet it bills at the same 5 credits per second as Gen-4 Turbo.
The same page lists Wan 3 at 5, 10 and 20 credits per second for 480p, 720p and 1080p. Wan 3 Prime runs from 6.8 to 28. Seedance 2.5 runs from 20 to 68 plus input and reference fees, Seedance 2.0 from 36 to 150, and Grok Imagine Video 1.5 from 10 to 29 plus 1 credit per reference.
Video is not the only line item. Smaller charges add up in a pipeline that makes stills, voiceover and final upscales:
The minimum purchase of 1,000 credits stretches differently on each model:
| Model | Footage for 1,000 credits | Five-second clips |
|---|---|---|
| Gen-4 Turbo | 200 seconds | 40 |
| Gen-4.5 | about 83 seconds | 16 |
| Veo 3.1 Fast with audio | about 66 seconds | 13 |
| Veo 3.1 with audio | 25 seconds | 5 |
The ratios matter more than the dollars. Gen-4.5 costs 2.4 times what Gen-4 Turbo costs per second, and Veo 3.1 with audio costs 8 times as much. Switching audio off on either Veo tier cuts the rate by a third or a half, depending on the tier.
Setup is short, but one step is easy to botch: the secret string is shown only once.
blog-video-prod.
💡 Create separate credentials for development and production. If a test secret leaks, you can revoke it without breaking the live app.
The official SDKs read the secret from an environment variable named RUNWAYML_API_SECRET, so you never paste it into source code.
On macOS and Linux:
export RUNWAYML_API_SECRET="your_secret_goes_here"
On Windows:
setx RUNWAYML_API_SECRET "your_secret_goes_here"
On Windows, setx only affects terminals opened afterward, so restart your shell before testing. Then follow four habits:
The short answer: the official pricing documentation lists no free API tier. Anything free sits next to the API, not inside it.
The pricing page lists rates and a $10 starting payment, with no complimentary API balance and no trial credits. Runway's web app has a free plan, listed at 125 one-time credits, but that balance lives in the web workspace and cannot be spent on API calls. The cheapest real test of the API is therefore the $10 minimum: 1,000 credits, or 40 five-second clips on Gen-4 Turbo.
If you only want to see what a model can do, skip the billing step. PicassoIA lists Seedance 2.5 Lite as a free, unlimited generator for clips up to ten seconds, and Picasso IA Video as a free, unlimited generator from text or an image. Runway's own models run in the browser there too: Gen-4.5, Gen-4 Turbo, Aleph 2 for restyling a full video from one edited frame, and Gen 4 Aleph for recutting footage. There is no developer account and no secret to manage.
| Option | Entry cost | Best for |
|---|---|---|
| Runway API | $10 minimum purchase | Apps and automated pipelines |
| Runway web free plan | 125 one-time credits | Manual testing in the web app only |
| Seedance 2.5 Lite | Free and unlimited per its model page | Drafts and short clips |
| Gen-4.5 on PicassoIA | See the model page for your plan | No-code Runway output |
Every project sits in one of five tiers. The tier sets how many generations can run at once, how many you can start per day, and how much you can spend per month.
| Tier | Concurrency | Daily generations | Monthly spend cap | Upgrade after |
|---|---|---|---|---|
| 1 | 1 to 2 | 50 to 200 | $100 | Base tier |
| 2 | 3 | 500 to 1,000 | $500 | $50 purchased |
| 3 | 5 | 1,000 to 2,000 | $2,000 | $100 purchased |
| 4 | 10 | 5,000 to 10,000 | $20,000 | $1,000 purchased |
| 5 | 20 | 25,000 to 30,000 | $100,000 | $5,000 purchased |
Do the math before you buy. Tier 1's $100 monthly cap equals 2,000 seconds on Gen-4 Turbo or about 833 seconds on Gen-4.5.
Most solo projects live happily in Tier 1. Concurrency of 1 to 2 means clips queue instead of failing, which suits batch jobs but feels slow in a live product. If your app shows a user a clip within moments of a click, plan to reach Tier 2 or Tier 3 ($50 or $100 purchased) before launch, since each step raises the number of generations that run at the same time.
Hitting the concurrency limit does not return an error. New tasks receive a THROTTLED status and wait in the queue in the order you submitted them. There is no requests-per-minute cap, but the daily limit runs on a rolling 24-hour window rather than resetting at midnight. Upgrades apply automatically once your purchases reach the next threshold, with no waiting period, and teams that need guaranteed concurrency can file an exception request in the portal.
Published rates describe a perfect generation. Real work involves retries, so budget for attempts, not for finished shots.
These scenarios are illustrative, and the attempt counts are assumptions you should replace with your own hit rate.
| Scenario | Model | Generations | Credits | Cost |
|---|---|---|---|---|
| 30 social clips, 5 seconds each, 3 attempts per clip | Gen-4 Turbo | 90 | 2,250 | $22.50 |
| 12 hero shots, 10 seconds each, 4 attempts per shot | Gen-4.5 | 48 | 5,760 | $57.60 |
| 10 ad spots with sound, 5 seconds each, 3 attempts | Veo 3.1 with audio | 30 | 6,000 | $60.00 |
All three fit under the Tier 1 cap of $100 per month, though the second and third would use most of it.
Three mistakes inflate bills more than anything else. The first is testing prompts on the priciest model. The second is forgetting that retries multiply every rate in the tables above. The third is running an unattended loop with no spend ceiling, so set a monthly limit in your own code that sits below your tier cap.
Choosing between the API and a subscription comes down to volume and automation. If a person clicks Generate a few times a week, a web plan or a browser generator is simpler. If code triggers generations, or you need predictable per-clip billing across many users, the API's pay-per-second model is the cleaner fit.
When you want Runway-grade output without code, the browser route is faster than building an integration. Gen-4.5 is listed as cinematic text-to-video.
💡 Sample prompt: A barista slides a flat white across a wooden counter, slow dolly in, soft window light from the left, shallow depth of field, natural film grain.
You now have the numbers: $0.01 per credit, 5 to 40 credits per second depending on the model, a $10 starting purchase, and five tiers that cap spend and concurrency. Decide what a finished shot is worth to you, divide by your attempt count, and the right model usually picks itself.
A quick rule of thumb:
If the API is more setup than your project needs, open Picasso IA and make your own videos and images in the browser. Start with a free generator such as Seedance 2.5 Lite, compare it with Gen-4.5, and keep the one that fits your budget. Browse every model at picassoia.com/en/all-models, write your first prompt, and see how far a single idea can go.
