Runway makes some of the most controllable AI video tools around. Runway says Gen-4.5 took the top spot on the Artificial Analysis text-to-video leaderboard when it launched in late 2025, though leaderboards move, so check the live ranking rather than taking that as today's position. This guide covers what the API does, what it costs, how to get access, and a working code example to generate your first clip.
What the Runway API Is
The Runway API gives developers direct access to Runway's video models, currently Gen-4 Turbo and Gen-4.5, without going through the consumer web app. It handles both text-to-video and image-to-video: give it a prompt on its own, or a prompt plus a starting image, and it builds a clip from there.
The API is video-focused. There is no image generation endpoint attached to these two models the way there is with GPT Image or Nano Banana, and Runway does not generate songs the way Suno does. It does offer separate audio endpoints for text-to-speech and sound effects, just not music. If you need image generation or music generation, you will be pairing Runway with something else.
Features
Text-to-video and image-to-video. Both models take a prompt alone or a prompt with a reference image that becomes the opening frame. Image-to-video is the more dependable route when a particular subject, character, or scene has to stay the same, since the model starts from something concrete instead of inventing it.
Camera movement and a consistent look. Runway's reputation for control comes mostly from the web app, where Motion Brush lets you paint over parts of an image to direct movement, and separate camera controls set pans, arcs, and zooms. Through the API, that control comes from how you word the prompt rather than from dedicated settings. Camera moves described in words, such as a slow push toward the subject or an arc around it, come through reliably. Motion Brush does not: it is a web-app feature with no API equivalent today. For holding a character steady across several clips, our guide to keeping AI-generated characters consistent covers the techniques that do work through an API.
Resolution, length, and shape. Both models output 720p in 5 or 10 second clips. Aspect ratios cover 16:9, 9:16, 1:1, 4:3, 3:4, and 21:9, so you can target landscape, portrait, or square without cropping afterward. Two limits worth weighing: 720p is below what Veo, Kling, and Seedance now offer, and neither Runway model produces sound. Most rivals now generate audio in the same pass, so if your clip needs it you are adding a second step with something like ElevenLabs. Our breakdown of which APIs generate synchronized audio covers who does what.
Pricing
How Runway's own credits work
Runway's developer API sits on a separate site, dev.runwayml.com, with a credit pool entirely separate from consumer Standard, Pro, or Max subscriptions. Credits are $0.01 each and you need a minimum $10 top-up before your first call. No waitlist, no approval step: sign up, add credits, start.
Cost per second by model
Going direct, Gen-4 Turbo is 5 credits per second ($0.05), so a 5-second clip is $0.25 and a 10-second clip is $0.50. Gen-4.5, the flagship, is 12 credits per second ($0.12), so $0.60 and $1.20 for the same lengths.
Why other providers charge more
Platforms that carry Runway alongside other video, image, and music models usually charge more per second than Runway does. You are paying for one integration and one bill across many models instead of managing each provider separately. On Apiframe, Gen-4 Turbo is 36 credits for 5 seconds or 73 for 10 (about $0.36 and $0.73), and Gen-4.5 is 87 credits for 5 seconds or 174 for 10 (about $0.87 and $1.74). That works out at roughly 1.4 times Runway's own rate on both models, so the question is whether one request format across the whole catalog is worth about 45 percent more per clip. If Runway is the only model you will ever call, go direct.
Other providers sit at different points on that same trade-off. Some flat-rate resellers undercut Runway's own API but support less, image-to-video only for instance, while others charge more for broader coverage. Worth working out your expected monthly spend across a few before committing, since the answer depends on whether you need Runway alone or a catalog behind one key.
Getting Access
Going direct
Runway's API is not behind a waitlist or a sales call, which surprises people expecting it to work like other closed video models. Three steps: create an account on the developer site, create an organization, which holds your keys and billing settings, then add the minimum $10 in credits.
Going through a platform
Using a platform that carries many models skips the Runway developer account entirely. You sign up once, take a key from the dashboard, and Runway sits alongside everything else on the same key and the same balance. Our getting started guide walks through that.
Keys and authentication
On Runway's portal you create a key under the API Keys tab, copy it straight away since it is shown once, and pass it through the RUNWAYML_API_SECRET environment variable, which Runway's official libraries read automatically.
On Apiframe you generate a key from the dashboard. Keys begin with afk_ and go in the X-API-Key header on every request. New accounts get 100 free credits, which covers two 5-second Gen-4 Turbo clips at 36 credits each, or one Gen-4.5 clip at 87, so enough to compare the two tiers before paying.
Code Example
Your first request
Here is a text-to-video request through Apiframe, one of the two routes described above.
import requests
response = requests.post(
"https://api.apiframe.ai/v2/videos/generate",
headers={
"X-API-Key": "afk_your_api_key_here",
"Content-Type": "application/json",
},
json={
"prompt": "a cinematic slow-motion shot of ocean waves crashing on volcanic rocks",
"model": "runway-gen4.5",
"runwayParams": {
"duration": 10,
"aspect_ratio": "16:9",
},
},
)
data = response.json()
print(data) # {"jobId": "...", "status": "QUEUED"}That 10-second Gen-4.5 request costs $1.74, so switch "model": "runway-gen4.5" to "runway-gen4-turbo" while you are testing to spend less than half that. For image-to-video, add an image field inside runwayParams pointing at a publicly reachable image URL, and the model uses it as the opening frame.
Waiting for the result and handling errors
Generation runs in the background. The request above returns a job ID immediately and you check a separate endpoint for progress.
import time
import requests
job_id = data["jobId"]
headers = {"X-API-Key": "afk_your_api_key_here"}
while True:
job = requests.get(f"https://api.apiframe.ai/v2/jobs/{job_id}", headers=headers).json()
status = job["status"]
if status == "COMPLETED":
print(job["result"]["videoUrl"])
break
elif status == "FAILED":
print("Generation failed:", job.get("error"))
break
time.sleep(5)Jobs move through QUEUED, then PROCESSING with a progress percentage you can show users, then COMPLETED or FAILED. Failed jobs are refunded, so a failure costs nothing. Handle 402 for running out of credits, 429 for too many requests, 400 for a bad runwayParams value, and 401 for a missing or wrong key. Rather than checking in a loop, both Runway and most platforms can call your server when a job finishes, which our walkthrough on callbacks covers.
Timeouts need separate handling from the codes above. A timeout means the reply never arrived, not that the job never started, so resending blindly can run a second generation and bill you for both. At $1.74 for a 10-second Gen-4.5 clip that adds up quickly. Send an idempotency key so the platform recognizes the repeat. Finished videos are hosted for 90 days, so download anything you want to keep.
How It Compares
Prices below are per second at Apiframe's standard tier, so Runway can be compared against other models on the same basis. Each provider's own direct pricing will differ, and Runway's is lower as noted above.
| Model | Cost per second | Max clip | Resolution | Generates audio |
|---|---|---|---|---|
| Runway Gen-4 Turbo | ~$0.073 | 10s | 720p | No |
| Runway Gen-4.5 | ~$0.17 | 10s | 720p | No |
| Kling 3.0 (standard) | ~$0.24 | 15s | Up to 4K, higher tiers cost more | Yes |
| Veo 3 / 3.1 | ~$0.29 silent, ~$0.58 with audio | 8s | Up to 1080p | Yes |
| Seedance 2.5 | Varies by resolution | 30s | Up to 720p | Yes |
| Hailuo 2.3 | Flat per clip, from $0.41 | 10s | Up to 1080p | No |
Runway's edge is not price, and it is not specs either, since 720p and 10 seconds now trail the field. It is control: camera movement you can direct through the prompt and a look that holds from shot to shot, which makes it a common pick for film-adjacent and branded work where hitting a specific feel matters more than cost per clip. If the lowest cost per second is what you need, or sound in the same pass, Kling 3.0, Veo 3, and Seedance 2.5 all beat it there. Our AI video API guide covers the wider field.
Sora is absent from that table on purpose: OpenAI shut its API down on September 24, 2026, so it is no longer something to build on. Runway is one of the places teams moved, as our roundup of Sora alternatives covers.
FAQ
Is there a free Runway API tier?
Not on Runway's own API, which requires the $10 minimum top-up before your first call. Through Apiframe, new accounts start with 100 free credits, enough for two 5-second Gen-4 Turbo clips or one on Gen-4.5.
What are the rate limits?
Runway documents its own limits on the developer portal, and they vary by account tier, so check your dashboard for how many requests you can run at once. Apiframe allows 500 requests per minute per account across every model, not just Runway. For video work, the cap on simultaneous jobs usually bites before the request limit does.
Can I use the output commercially?
Both Runway directly and third-party access allow commercial use under their respective terms, and if you go through a platform, its terms apply on top of Runway's. Read the current ones before shipping at volume, since rules on redistribution and training data change.
How long does a generation take?
There is no fixed figure. Runway publishes no guaranteed turnaround, and real time depends on model, length, resolution, and how busy the queue is. Check the job status or use a callback rather than assuming a wait, and build your interface around a processing state instead of a countdown.
Should I go direct or through a platform?
Direct, if Runway is the only model you need. Setup takes minutes with no waitlist and you avoid the roughly 45 percent markup. Through a platform, if you are already calling other video or image models, or expect to, since one request format and one bill across the catalog usually outweighs that difference once you are maintaining several integrations. For a fuller comparison of where to get Runway, including providers beyond Apiframe, see our Best Runway API Providers Compared guide. You can also see how Runway places in our Best AI Video Generation APIs roundup, or how prices compare across models in AI Video API Pricing in 2026.
Apiframe Team
The team behind Apiframe - making AI generation accessible to everyone.