Runway Gen-4.5 API and Pricing, Explained Through a Real Pipeline

Runway Gen-4.5 is the model that carries Runway's filmmaker reputation into a pipeline, and its price is the simplest in our catalog. One rate per second of output. No resolution tiers, because the endpoint renders one resolution. No audio surcharge, because the model renders picture only. The cost of a clip is the rate times the seconds, and the length is the only decision.

This post covers what Gen-4.5 costs, what the endpoint accepts and refuses, why your balance reserves more than a clip costs, and how to call the model as a published API endpoint. Every parameter and price below is read off Treza's live video model catalog as of September 2026.

What you get before you talk about price

Two rows stand out. The 2 second floor is lower than most of the lineup and gets its own section below. The absence of audio changes what the graph around the node has to do, not what the node can do.

What Gen-4.5 is for, in our own editorial positioning rather than a benchmark, is scene craft: framing, light, and motion that read like coverage from a set, with prompt adherence that keeps the shot you asked for. That matters most when a storyboard, a client, or an upstream script step decides the shot and the render has to honour it. Full capability detail lives on the Runway Gen-4.5 model page.

How Gen-4.5 bills

Per second of output, at one flat rate:

The catalog lists Gen-4.5 at 12 cents per second of output, and that is the whole rate card. There is no cheaper draft resolution and no premium one, no audio switch, and no discount at the top of the range. A 10 second take costs exactly five times a 2 second one.

For scale, Kling 3.0 Pro is 11.2 cents per second silent and 16.8 with audio, and FLUX.3 Video is 17 cents at 720p. For a silent 720p shot, the shot most narrated pipelines need, Gen-4.5 sits at the low end of that group.

The 2 second floor is the interesting row

Most video models in the catalog start at 4 or 5 seconds. Veo 3.1 renders exactly 4, 6, or 8. Kling Video O1 renders 5 or 10. If the shot you need is a whip pan, a one-beat insert, or a B-roll stab under a single line of narration, those grids make you pay for seconds you will cut.

Gen-4.5 renders 2 seconds for 24 cents of provider cost, and every whole second above that up to 10. Inserts and transitions stop being something you trim out of a longer clip and become something you generate at the length you need. A narrated Short that needs five cutaways can render all five on Gen-4.5 for $1.20 combined, the price of one 10 second clip. The same grid lets a shot match the narration line it sits under, so the Sequence node has less to pad and less to cut.

What you actually pay on Treza

Gen-4.5 through a Treza pipeline is metered from a prepaid credit balance. One credit equals $0.01, credit packs start at $5, credits never expire, and there is no subscription. A failed generation charges nothing.

One detail is worth knowing before your first run. The platform estimates the media cost of a graph and refuses to start a run your balance cannot cover. Gen-4.5 has no entry in the estimator's table of verified per-second rates, so it falls back to a conservative floor of 38 cents per second with a $2.26 minimum per clip. A 10 second node holds $3.80 against a published cost of $1.20, and a 2 second node holds the $2.26 minimum against 24 cents.

That hold is not a price. The run settles at actual provider cost plus markup and the difference returns to the balance. The floor errs high on purpose, so a model the estimator does not recognise can never start a job the balance cannot finish. The gap is widest on exactly the short inserts Gen-4.5 is good at, so a graph with several 2 second nodes needs a balance that covers the holds even though the settled charge will be a fraction of them.

For budgeting across the catalog, the figure on the pricing page is $1.06, the median charge for one video generation. The cost breakdown post walks a finished video end to end.

What the endpoint will and will not take

On the endpoint we run, you send a prompt, a duration, an aspect ratio, and optionally a first frame image. Give it a first frame and it animates out of that still, which is how you open on a product shot or on a composition somebody has already approved. The first frame technique post covers checking that still before you pay for the video, and the image to video generator covers the job from the still's side.

It does not take a last frame. Kling 3.0 Pro and FLUX.3 Video accept both keyframes and generate the motion between them; on Gen-4.5, where the shot lands is up to the prompt.

It does not accept multi-image reference inputs, the kind that carry a character's identity from separate photos into a new shot. If you need that, Seedance 2.5, Seedance 2.0, and Wan 2.7 take them. With Gen-4.5, consistency across shots comes from the first frame: generate the still once, then hand it to each shot that opens on it.

It renders 720p in landscape or portrait. For phone-first vertical work that costs nothing. For a hero shot shown large, run the clip through the Video Upscale node, or render that shot on Veo 3.1, which reaches 4K.

Gen-4.5 supports seeds, so a shot you like can be regenerated reproducibly while you vary one thing at a time. On the canvas, the node's Result output returns the full provider response for each render, including the seed it used.

Picture only is a pipeline job, not a limitation

Gen-4.5 renders no sound. On a single-model tool that is a gap you fill by hand. On a pipeline it is one fewer thing to switch off, because the audio in a narrated piece was never going to come from the video model anyway.

The pattern we run: Gen-4.5 for the footage, a Text to Speech node for the narration, a music node for the bed, and the Sequence node to mix them into one file. The Sequence node stitches up to 12 shots, and the timeline editor exports up to 40 clips or 15 minutes. Compare that with a model whose native audio you pay for and then replace: on Kling 3.0 Pro the audio switch alone is a third of the bill on every clip that has it on.

Calling Gen-4.5 as an API without building the integration

Direct API access to a video model is never one HTTP call. It is credentials, a submit-and-poll job pattern, retries, moderation rejections, and somewhere to store the output.

On a Treza AI video pipeline, Gen-4.5 is a node on a canvas. Set the duration and aspect ratio, wire it next to a script step, a voiceover step, or a captioning step, and publish the whole chain as one versioned endpoint. Your application calls a typed /invoke route with JSON in and JSON out, or points an existing OpenAI SDK at the OpenAI-compatible route.

The swap is what pays off later. When the next model wins the leaderboard, you change the dropdown and republish, and the endpoint your product calls does not change. Check that the new model supports your node's duration, since a 2 or 3 second shot is below the floor of most of the catalog. Our hosted MCP server exposes the same pipelines as tools, so an agent can render a brief on Gen-4.5 and return the link.

Where Gen-4.5 earns its place in a graph

Three jobs, from our own pipelines. Storyboarded hero beats, where the shot was decided upstream and the render has to match it, which is where prompt adherence saves rerolls. Inserts and transitions in short vertical cuts, where a 2 second generation at 24 cents replaces a 5 second one you would trim. And silent footage under narration in a faceless video pipeline, where a picture-only model at 12 cents a second is the honest price for a shot whose sound was always going to be the voiceover.

The fastest way to answer what it costs for your own work is to run your real prompt at your real duration and read the charge. Open the AI video generator, pick Runway Gen-4.5, set the length the shot actually needs, and generate once.

Frequently Asked Questions

How much does Runway Gen-4.5 cost per second?

Twelve cents per second of output, as published for the endpoint we run as of September 2026. There are no resolution or audio tiers, so a 2 second clip is 24 cents of provider cost and a 10 second clip is $1.20. On Treza that is charged from a prepaid credit balance, and a failed generation charges nothing.

What is the longest clip Gen-4.5 can generate?

Ten seconds in a single generation, at any whole number of seconds from 2 to 10. For longer pieces, stitch generations with the Sequence node, which handles up to 12 shots, or assemble them in the timeline editor, which exports up to 40 clips or 15 minutes.

Does Runway Gen-4.5 generate audio?

No. It renders picture only, with no audio option on the endpoint we run. On a Treza pipeline the usual pattern is a Text to Speech node for narration, a music node for the bed, and the Sequence node to mix them with the footage into one file.

Can I give Gen-4.5 a first and last frame?

A first frame, yes. A last frame, no. Kling 3.0 Pro and FLUX.3 Video accept both keyframes if the shot has to land on an exact composition. Gen-4.5 also does not accept multi-image character references; Seedance 2.5, Seedance 2.0, and Wan 2.7 do.

Why does my credit balance reserve more than the clip costs?

Gen-4.5 is not in the estimator's table of verified per-second rates, so it falls back to a floor of 38 cents per second with a $2.26 minimum. A 10 second node holds $3.80 against a published cost of $1.20. The run settles at actual provider cost plus markup and the difference returns to your balance. Reserves are holds, not prices.

Can I call Runway Gen-4.5 through an API without a Runway account?

Yes. On a Treza pipeline, Gen-4.5 is a node you configure once and publish as a versioned HTTP endpoint, with a typed /invoke route and an OpenAI-compatible route. A hosted MCP server exposes the same pipelines as tools for agents. There is no subscription: generations are metered from prepaid credits that start at $5 and never expire.