
Runway Gen-4 Turbo API: Pricing, Specs, How to Access
Runway Gen-4 and Gen-4 Turbo: what the API and plans cost, 2-10s 720p specs, its Arena rank (49th of 50), and the video APIs you can call today instead.
Runway Gen-4 is the image-to-video model Runway released on March 31, 2025, and Gen-4 Turbo is its faster, lower-credit version, listed on Runway's API as gen4_turbo since April 4, 2025. Both turn one starting image into silent 720p-class video. In November 2026 Gen-4 Turbo is a legacy draft tier: Arena's image-to-video board ranks it 49th of 50, and Runway's own API now lists 29 video models, only four of them made by Runway.
TL;DR: Gen-4 Turbo is image-to-video only, takes 2 to 10 seconds per request on Runway's API, outputs at six fixed 720p-class sizes with no sound, and bills Runway credits per second (well under half the per-second credits of Gen-4.5). Base Gen-4 lives only in Runway's app. On today's Arena boards it scores 1054, more than 400 points behind the leaders. We would not start a new integration on it. Inside Runway, use Gen-4.5; outside, the closest live models on UnificAlly are Veo 3.1, Kling 3.0, SeeDance 2.0 and Wan 2.6, all with sound and all ranked far higher. Runway models are not available on UnificAlly.
What is Runway Gen-4?
Runway built Gen-4 around one problem: keeping the same character, object or place recognizable across separate generations from a single reference image, with no fine-tuning. Its launch page frames that as world consistency, with references and instructions combined to place a subject in new scenes. That is why the Runway Gen-4 AI video generator still draws searches in 2026, long after newer models passed it.
The model works only from an image. The picture you send becomes the first frame, and the prompt should describe motion rather than the scene. Searchers spell the name several ways (Runway Gen-4, Runway Gen 4, runway gen4), and the questions behind them repeat: Runway Gen 4 pricing and Runway Gen 4 credits, the Runway Gen 4 resolution, whether Runway Gen 4 image to video and Runway Gen 4 text to video both exist, whether Runway Gen 4 free use is possible, and Runway Gen 4 vs Gen 4.5. Each has its own section below.
| Spec | Gen-4 (app) | Gen-4 Turbo (gen4_turbo on the API) |
|---|---|---|
| Input | Image (required) + text prompt | Image (required, first frame only) + optional text prompt |
| Durations | 5 or 10 seconds in the app | Any whole number from 2 to 10 seconds |
| Output resolution | 720p class | 720p class |
| Aspect ratios | 16:9 (1280×720), 9:16 (720×1280), 1:1 (960×960), 4:3 (1104×832), 3:4 (832×1104), 21:9 (1584×672) | Same six, sent as ratio strings such as 1280:720 |
| Frame rate | 24fps | 24fps |
| Prompt limit | 1,000 characters | 1,000 UTF-16 code units |
| Sound | None | None |
Around the edges: a finished video can be upscaled to 4K as a separate action in the app, a fixed seed (0 to 4,294,967,295) reproduces similar motion, and audio is a separate tool in Runway's suite. Gen-4 References covers consistent characters in still images (API model gen4_image, with gen4_image_turbo as the fast version), and Act-Two (act_two) drives a character with your own recorded performance.
Runway Gen-4 Turbo release date and features: where it stands in 2026
Gen-4 came out on March 31, 2025, and Runway Gen 4 Turbo followed days later; Runway's API model table dates gen4_turbo to April 4, 2025. Nothing new has shipped for Gen-4 Turbo in 2026. Runway's help center files Gen-4 under an older generation, and Runway's 2026 API additions have been Gen-4.5's options, Aleph 2.0 for editing and a long list of other labs' models.
What Gen-4 Turbo still does, per Runway's API reference:
- One input image, first frame only.
promptImageaccepts onlyposition: "first"; there is no last-frame option. - Prompt optional.
promptTextcan be left out and the model animates the image on its own; with it, describe the motion in up to 1,000 characters. - 2 to 10 seconds.
durationis an integer in that range, not just the 5 or 10 the app offers. - Six fixed sizes.
ratiomust be one of1280:720,720:1280,1104:832,832:1104,960:960or1584:672. - Moderation switch.
contentModeration.publicFigureThresholdset tolowrelaxes the public-figure check;autois the alternative. - What it lacks against 2026 models: no text-to-video, no sound, no reference images beyond the first frame, no 4K without a separate upscale.
Runway Gen-4 Turbo vs Veo 3.1 vs Kling 4.0
The two pay-per-use video APIs most often weighed against Gen-4 Turbo. Runway's column comes from its API reference; the other two are live on UnificAlly at our rates.
| Runway Gen-4 Turbo | Veo 3.1 Lite Relaxed | Kling 4.0 | |
|---|---|---|---|
| Billing | Runway credits per second | $0.075 flat per 4 to 8 s video at 720p | $0.126 per second (std) |
| 5 second image-to-video | Credits, see runway.com/pricing | $0.075 (4 or 6 s; 5 s is not offered) | $0.63 |
| Text-to-video | No | Yes | Yes |
| Durations | 2 to 10 s | 4, 6 or 8 s, extendable | 3 to 30 s in one call |
| Resolution | 720p class, 4K via app upscale | 720p, 1080p, 4K | 720p (std) or 1080p (pro) |
| Native audio | No | Yes | Yes, stereo |
| References | First frame only | Up to 3 reference images, or start and end frame | Start frame, elements, keyframes |
| Available on UnificAlly | No | Yes | Yes |
What the alternative gives you: a real image-to-video test
To make the comparison concrete we ran a Gen-4-style job, one product image plus a motion prompt, on Veo 3.1 Lite Relaxed through the UnificAlly MCP on October 19, 2026. The start frame came from Grok Imagine 1.0 Image ($0.01); the video cost $0.075.
- Start frame prompt (Grok Imagine 1.0 Image, 16:9): "Product photo of a matte black wireless over-ear headphone on a walnut desk beside a steaming ceramic mug, warm morning window light from the left, shallow depth of field, photorealistic".
- Video prompt (Veo 3.1 Lite Relaxed, 720p, 4 s, 16:9, seed 66363): "Slow dolly-in on the headphones as steam rises from the mug, morning light shifts gently across the desk, the headphones rotate a few degrees on their stand. Cinematic product shot. AUDIO: quiet room tone, soft ceramic clink."
Across the four seconds the steam keeps moving and the window light travels over the desk, while the product, desk and framing stay exactly as in the still: the first-frame fidelity Gen-4 made its name on, plus a soundtrack. The dolly-in is barely visible and the headphones never rotate, so Lite Relaxed treated the hero object as fixed. For a product shot that is the safe way to fail; for a turntable move, make the rotation the only motion in the prompt and step up to Fast.
How good is Gen-4 Turbo now? Arena puts it 49th of 50
Arena's crowd-voted image-to-video board, dated November 7, 2026 and checked November 16, ranks Runway Gen-4 Turbo #49 of 50 at 1054 ± 13 over 6,801 votes. Only Pika 2.2 sits below it. The live models listed in the alternatives section are all far higher on the same board:
| Image-to-video, Arena (Nov 7, 2026) | Rank | Score |
|---|---|---|
| Seedance 2.0 (720p) | #6 | 1480 ± 7 |
| Veo 3.1 (Audio) | #15 | 1399 ± 10 |
| Kling 3.0 Pro | #21 | 1353 ± 5 |
| Wan 2.6 | #25 | 1309 ± 7 |
| Runway Gen-4 Turbo | #49 | 1054 ± 13 |
Gen-4.5, Runway's current model, is not on the image-to-video board. On the text-to-video board it is #29 of 50 at 1223 ± 8, level with Kling 2.6 Pro (1216 ± 7). A gap of 250 or more points to every alternative in the table is the reason for our call: Gen-4 Turbo is fine for cheap motion tests inside a Runway workflow you already have, and a poor base for new work.
Runway Gen 4 vs Gen 4.5: is Gen-4 still Runway's latest model?
No. Runway announced Gen-4.5 on December 1, 2025. Its launch post claimed first place on the Artificial Analysis text-to-video leaderboard at 1,247 Elo, measured on November 30, 2025; a year later, Arena's text-to-video board has it at #29. Runway's help center now places Gen-4 in its older generation.
What changes at the request level, from Runway's API reference:
- Runway Gen 4.5 does text-to-video. No input image needed, 16:9 or 9:16 (
1280:720,720:1280). Image-to-video uses the same six sizes as Gen-4 Turbo. - Prompt and duration become required. Gen-4.5 needs
promptTextand an explicitdurationfrom 2 to 10 seconds; Gen-4 Turbo makes both optional. - Delivery formats. Gen-4.5 takes an
outputFormatsuch asprores,png_sequence,hdr10orhlg(ProRes profiles from 422 Proxy to 4444 XQ). Gen-4 Turbo returns MP4 only. - Same app credits as base Gen-4. Runway's plan page counts the same seconds of Gen-4.5 and Gen-4 per credit allowance; Standard or a higher plan is needed for Gen-4.5 in the app.
So why use Gen-4 at all? Gen-4 Turbo costs well under half of Gen-4.5's per-second credits, and Runway still recommends it for testing an idea before spending on a final pass. For video-to-video editing, Runway's current answer is Aleph 2.0. On UnificAlly, the editing models are HappyHorse 1.0 Video Edit for reference-image swaps and restyles, and the Kling O1 Edit API when the aspect ratio has to change too.
Runway pricing plans and Gen-4 credits
Runway bills everything in credits. In the app, credits come with a subscription; on the developer API, you buy them directly at a flat rate per credit, with no plan needed.
The Runway pricing plans are Free (a one-time credit grant that does not refill), Standard, Pro, Max and Enterprise; current prices and monthly allowances are on runway.com/pricing. Extra credits can be bought on any paid plan. Per second, base Gen-4 and Gen-4.5 cost the same number of credits in the app, and Gen-4 Turbo costs well under half of that. Because credits are spent per second of output, a 10-second video costs twice a 5-second one on every model, and on the API a 2-second Gen-4 Turbo test costs a fifth of a 10-second one.
Runway API pricing and how to call Gen-4 Turbo
Yes, Runway has an API. The developer platform lives at dev.runwayml.com, with docs at docs.dev.runwayml.com; Runway API pricing is in credits per second (per image for image models), listed on the docs' pricing page. Base Gen-4 is app-only, so on the API the Gen-4 route is gen4_turbo.
By November 2026 the Runway API is mostly a catalogue of other labs' models. Its model index lists 29 video model IDs, and only four are Runway's own:
| Runway-made API model | Task | Billed |
|---|---|---|
gen4_turbo | Image-to-video | Per second |
gen4.5 | Text- and image-to-video | Per second |
act_two | Performance capture | Per second |
aleph2 | Video-to-video editing | Per second |
gen4_image | Image generation with references | Per image, by resolution |
gen4_image_turbo | Fast image generation | Per image |
The other 25 video IDs include Kling 4.0 and 4.0 Flash, Kling 3.0 and O3, Seedance 2.5 and the Seedance 2.0 family, MiniMax H3, Wan 3.0, Grok Imagine Video 1.5, HappyHorse 1.0, Veo 3.1 and Gemini Omni Flash, plus eight ElevenLabs audio models.
Runway publishes two SDKs, runwayml on PyPI and @runwayml/sdk on npm. A Gen-4 Turbo call in Python:
from runwayml import RunwayML
client = RunwayML() # reads RUNWAYML_API_SECRET
task = client.image_to_video.create(
model="gen4_turbo",
prompt_image="https://example.com/first-frame.jpg",
prompt_text="Slow dolly-in as the subject turns toward the window",
ratio="1280:720",
duration=5,
).wait_for_task_output()
print(task.output[0])
Behind the helper: every raw HTTP request carries the header X-Runway-Version: 2024-11-06, and a task moves through PENDING, THROTTLED (queued behind your rate limit, not an error), RUNNING and then SUCCEEDED, FAILED or CANCELLED. Only a succeeded task has the output array of video URLs, and those URLs are temporary, so download the file rather than storing the link. API keys are issued in the developer portal after sign-up.
Runway Gen-4 alternatives you can call on UnificAlly today
Runway's models are not on UnificAlly, and we are not going to pretend otherwise. To call Gen-4 Turbo or Gen-4.5, use Runway's own API or app as described above.
For image-to-video billed per request, with no plan and no credit packs, these are the closest models live on UnificAlly, each ranked far above Gen-4 Turbo on Arena's image-to-video board:
- Veo 3.1: 4, 6 or 8 seconds, 720p to 4K, with dialogue, ambience and effects generated in the same call, plus up to 3 reference images or a start and end frame. From $0.075 per video. Gen-4 is silent, so sound is the clearest gap.
- Kling 3.0: 3 to 15 seconds, multi-shot with 2 to 6 scenes per call, a 4K mode and optional sound. From $0.0756 per second.
- SeeDance 2.0: up to 9 image, video and audio references against Gen-4's single input image, 4 to 15 seconds, up to 4K on Pro, with Mini and Fast at 480p or 720p. The Mini variant starts at $0.011 per second.
- Wan 2.6: text-, image- or reference-to-video with up to 5 references (at most 3 of them videos), 720p or 1080p, 2 to 15 seconds, with sound on by default.
All four run through one API key with pay-per-use billing and no monthly fee. New accounts start with a $0.50 balance; on video models that covers only a few seconds, so treat it as a smoke test rather than a trial. UnificAlly pricing shows the current rate for each of the four, and the full catalog is at /models.
Runway Gen-4 Turbo FAQ
How much does Runway Gen-4 cost?
Runway bills Gen-4 in credits per second of output, and Gen-4 Turbo costs well under half as many credits per second as base Gen-4. In the app, credits come with a monthly plan; on Runway's API you buy credits directly, and only Gen-4 Turbo is available there. Current prices are on runway.com/pricing.
Is Runway Gen-4 free?
Partly. Runway's Free plan gives a one-time grant of credits, enough for a few short Gen-4 Turbo videos, and it does not refill monthly. After that you need a paid plan or API credits.
What is the difference between Gen-4 and Gen-4 Turbo?
Same task and output sizes; Turbo generates faster and costs well under half the credits per second. Runway suggests drafting in Turbo and moving to Gen-4 only when Turbo's result is not good enough. The API offers only Turbo, as gen4_turbo, with any duration from 2 to 10 seconds.
Does Runway Gen-4 do text-to-video?
No. Gen-4 and Gen-4 Turbo need an input image, which becomes the first frame; the prompt describes the motion and is optional on the API. For text-to-video from Runway, use Gen-4.5, which generates 16:9 or 9:16 720p video from a prompt alone.
What resolution is Runway Gen-4?
Runway Gen-4 outputs 720p-class video in six sizes: 1280×720, 720×1280, 960×960, 1104×832, 832×1104 and 1584×672, all at 24fps. The app can upscale a finished video to 4K as a separate step; the API returns the 720p-class file.
Was Runway Gen-4 Turbo updated in 2026?
No. Gen-4 Turbo is the April 2025 model, unchanged, still on the API as gen4_turbo and still the draft model in the app. Runway's 2026 work went into Gen-4.5, Aleph 2.0 and hosting other labs' models, 25 of the 29 video IDs on its API.
How good is Runway Gen-4 in 2026?
Runway Gen-4 Turbo ranks #49 of 50 on Arena's image-to-video board (1054 ± 13, November 7, 2026). It keeps a subject consistent from one image, but it has no text-to-video, no sound and 720p-class output. For image-to-video with sound today, Veo 3.1 and Kling 4.0 do it pay-per-use.
Which AI video model is the most realistic?
By crowd vote on Arena's boards dated November 7, 2026, Gemini Omni 1.2 Flash leads both text-to-video (1540) and image-to-video (1503), each still marked preliminary. Seedance 2.5 and MiniMax H3 follow on image-to-video. Rankings move monthly, so run your own prompt on two or three models before choosing.
Can I use the Runway API without a subscription?
Yes. Runway's developer API at dev.runwayml.com bills its own credits, bought directly, with no plan required. Subscriptions only cover the app at runway.com.
If Runway reprices API credits, adds the extra Gen-4.5 input types it has promised, or retires gen4_turbo, the pricing, Gen-4.5 and API sections of this post change with it.


