Runway for Client Work: Prompts, Licensing and Delivery

TL;DR: Search "Runway commercial use" and you'll find confident answers with no citations. Runway's Terms of Use say it doesn't claim ownership of your outputs and doesn't restrict commercial use of them, on every plan including Free, with no credit required. What actually trips up agencies: the AI-training license in section 4.4, and outputs that aren't guaranteed unique.

Can You Actually Use Runway Video for Client Work?

If you're producing video for a client and the deliverable came out of Runway rather than one of the other AI video prompt tools agencies compare it against, the question that matters before you invoice isn't does it look good, it's am I allowed to sell this. Runway answers that directly, in its own Terms of Use, last updated May 11, 2026 (runway.com/terms-of-use, read September 4, 2026):

"The Company does not claim ownership of any of your Inputs or Outputs. Subject to your compliance with the Agreement, the Company does not restrict your commercial use of your Outputs."

That's section 4.4, User Inputs and Outputs, and it is not gated by plan. Runway's Help Center makes the same point in plainer language, in an article titled "Usage rights" that was itself updated September 2, 2026, two days before this was written:

"Yes, the content you create using Runway is yours to use without any non-commercial restrictions from us."

The same article lists exactly the use cases an agency cares about: "monetized or non-monetized YouTube uploads, other short-form video uploads (e.g. Instagram Reels, TikTok), social media posts, film festival entries, product advertising, and anything else that may require commercial use rights." A near-identical, slightly older article (help.runwayml.com, "Can I use the content I made in Runway for commercial purposes?", last updated July 30, 2025) says the same thing, so this isn't a one-off support answer; it's Runway's consistent published position across two articles a year apart.

Reselling the video itself, rather than just handing it to one client, is a separate question the ToS answers less directly, and it's worth reading the right clause rather than the wrong one. Section 1.3 of the Terms of Use restricts you from reselling Runway's Services: "you shall not license, sell, rent, lease, transfer, assign, reproduce, distribute, host, or similarly exploit any Services except as permitted through the APIs". That clause is about the platform, not your generated file; it stops you from reselling access to Runway itself, for example running your own paid "AI video generator" on top of someone else's Runway account. Selling your own Output, including through a stock-footage marketplace, falls under section 4.4's commercial-use grant instead, which names no restriction on the channel you sell it through.

This is not legal advice. It's a read of Runway's published terms as of the dates cited above, and Runway can change those terms at any time. If a contract with a client hinges on this, read the current page yourself, and if the deal is large enough to matter, have someone qualified read it too.

Do You Need to Credit Runway on Delivered Work?

No, and Runway says so directly. The same Usage Rights article states: "there is no formal requirement for you to credit us in your Runway-made work." It adds, separately, that Runway loves a shoutout but doesn't require one.

There's a real requirement hiding nearby that a lot of writing on this topic misses because it reads the wrong section. Runway's Terms of Use do require attribution, but only for a specific case: if you've purchased API access and built your own product or feature on top of it. Section 1 of the Terms of Use requires that such applications "must prominently display “Powered by Runway” and link to runway.com on applicable user interfaces". That's a rule for people embedding Runway's API inside their own software product, not for an agency generating a video in the Runway app and handing the client an .mp4. Confusing the two means either over-crediting a simple delivery or under-crediting an actual API integration.

The Clause Agencies Actually Need to Read: What Happens to Your Client's Uploads

Section 4.4 grants Runway more than a right to leave your outputs alone. The same clause that clears you to sell your outputs also has Runway keeping a license to use them:

This is where plan tier genuinely matters, and it's the part most is Runway safe for client work posts skip entirely. Runway's separate Enterprise Services Terms (runway.com/enterprise-terms, last updated June 1, 2026) draw a hard line the individual terms don't:

"Runway may not use Customer Content as training data for the Services."

"Customer Content" in that document is explicitly defined to include both Inputs and Outputs, and the same section states that "Customer owns all rights, title, and interest in Customer Content." There is no grant-back license to Runway for training purposes at all. Individual, Standard, Pro, Max and Team plans run on the consumer Terms of Use with the broader training license; only an Enterprise agreement removes it. If a client's procurement or legal team asks whether the vendor trains on their assets, the honest answer depends entirely on which contract you're actually under, not on how much you're paying per month.

Don't Promise "Exclusive" — Runway Won't Back That Claim

Agencies sometimes pitch AI video as bespoke and unrepeatable. Runway's own disclaimer section undercuts that pitch directly. From the Terms of Use, section 10, in Runway's own capitals:

"YOUR USE OF OUTPUTS IS AT YOUR SOLE RISK."

A few sentences later, the same section adds:

"DUE TO THE NATURE OF ARTIFICIAL INTELLIGENCE AND THE SERVICES, OUTPUTS MAY NOT BE UNIQUE AND SIMILAR OUTPUTS MAY BE GENERATED FOR OTHER USERS."

That same section also disclaims, in the same all-capitals style, any warranty of "MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT". Runway is not promising your generation is one of a kind, and it's not promising it doesn't accidentally resemble someone else's copyrighted style either. Runway's separate Usage Policy (runway.com/safety/usage-policy, last updated March 6, 2026) is worth reading for the second half of that risk: it lists "Content that may violate intellectual property rights" among prohibited uses, and separately prohibits "Attempts to create content in the style of a known, living artist". If a client brief asks you to chase a specific artist's or brand's look, that instruction, not just the output, is the thing that can put you on the wrong side of Runway's own policy.

None of this means don't use Runway for client work. It means write your contract around what Runway actually promises, which is commercial usage rights and nothing about exclusivity or originality.

Which Runway Plan Actually Fits Client Delivery?

Runway's published pricing (runway.com/pricing, read September 4, 2026) splits along lines that matter more for client work than the credit counts alone suggest:

Two things worth flagging precisely rather than guessing at. First, the commercial-use rights themselves aren't on this table because they don't vary by row: they come from the Terms of Use, and the Free plan's one-time 125 credits carry the same commercial rights as Max. Second, the pricing page never states outright that Free-tier output is watermarked; it only lists "No watermarks" as a feature that starts at Standard. Read literally, that's silent on Free rather than explicit about it, so if a client deliverable can't carry a visible mark, don't assume, generate the final pass on Standard or above, where "No watermarks" is a published feature rather than an inference.

For agencies running more than one client's work through a shared workspace, Team plans allocate credits per seat rather than to one account: Runway's own credits FAQ, updated September 2, 2026, lists the Team allotment as "6,900 per seat, shared across the workspace". That's a materially different budgeting model than upgrading one person's Pro plan and passing logins around, which the Terms of Use's account-sharing restrictions don't actually permit in the first place.

How Do You Prompt Gen-4.5 for a Client Brief, Not a Personal Project?

Once the licensing question is settled, the prompt itself is the easy part, provided you don't fight Runway's own guidance. Runway's Help Center, in the article "Introduction to Prompting" (updated June 27, 2026), answers a direct question about structured prompting: "Do JSON prompts provide more accurate results than natural language prompts?" Its answer: "JSON prompts give the placebo effect of being more accurate". A few sentences later, the same answer adds that "JSON formatting is ignored by generative models". Runway's own recommendation instead is to answer who, what, when, where and how in plain sentences. That's a claim about Runway's models specifically, not a universal rule; Google documents structured JSON prompting for its own Veo models, and this site's own JSON video prompt templates for Veo 3 reflect that vendor's documented approach, not Runway's.

For Gen-4.5 itself, verified directly from Runway's published OpenAPI spec (docs.dev.runwayml.com/openapi.json, read September 4, 2026), the actual request fields are narrower than most third-party guides describe:

• promptText: up to 1,000 characters, describing what should appear in the output.
• promptImage: on image-to-video, only a first frame position is supported. There's no last-frame or middle-frame slot on this model.
• ratio: differs by endpoint. Text-to-video's Gen-4.5 branch accepts exactly two values, 1280:720 and 720:1280. Image-to-video's Gen-4.5 branch accepts six: 1280:720, 720:1280, 1104:832, 960:960, 832:1104 and 1584:672.
• duration: an integer from 2 to 10 seconds.
• seed: an integer from 0 to 4294967295. Runway's own field description says using the same seed "will produce similar results" — not identical ones — so don't promise a client frame-exact reproducibility off a seed number alone.
• There is no negativePrompt field and no audio field on Gen-4.5 at all. If negative prompting is central to your workflow, that's worth knowing before you build a client-facing template around a field this model doesn't have; see this site's negative prompt support matrix for which vendors actually support it.

A plain-language template that fits Runway's own recommendation, built around who/what/when/where/how rather than a schema:

[Subject/who]: what appears in frame, with enough specificity that two people would picture the same thing
[Action/what]: what the subject is doing, as a single clear action
[Setting/where]: location, time of day, and any defining background detail
[Style/how]: lighting quality, lens feel, color palette — described in words, not a style-name shortcut
[Constraint]: anything that must NOT change between takes (brand color, product angle, on-screen text)

If you're running the same brief across several clients rather than one, don't rebuild this from scratch each time; swap the bracketed fields per client and keep the shape fixed instead.

What File Format Should You Actually Deliver to a Client?

Delivery format matters as much as the prompt once a client's editor is waiting on the file, and Gen-4.5's outputFormat parameter covers more of a real post-production pipeline than most third-party guides mention. Per Runway's own OpenAPI spec, it defaults to mp4 (H.264) and also accepts prores (a ProRes .mov), png_sequence (a zip of PNG frames, plus a .wav when the output has audio), sdr_rec709_10bit for SDR grading pipelines, and a set of HDR containers: hdr10 and hlg as 10-bit .mp4s, hdr_prores as a ProRes .mov editorial mezzanine, hdr_pq_12bit_master as a 12-bit mastering file, and hdr_png_sequence or hdr_exr_sequence for frame-sequence HDR delivery. A proresProfile parameter narrows the ProRes flavor, defaulting to 4444 for standard ProRes and 422 HQ for the HDR variant; Runway's own spec notes that 422 Proxy and 422 LT "quantize too heavily to hold the HDR gradients", which is why they aren't offered for hdr_prores.

None of this is free. Runway's spec states outright: "Non-mp4 formats incur an additional per-second credit surcharge: 5 credits per second for prores and png_sequence, and 20 credits per second for every 10-bit and deeper profile", rising to 40 credits per second once the frame is larger than 4 megapixels (past 1440p, up through 4K). If a client only needs a web-ready .mp4, requesting hdr_prores "for safety" is a real line item, not a free upgrade. Match the format to what the edit suite actually opens, and check the surcharge before you commit a shot list to anything other than the mp4 default.

A short checklist before a file leaves your machine for a client:

1. Confirm which Terms of Use actually govern the account you generated on: consumer terms (with the training-data license) or an Enterprise agreement (without it).
2. Regenerate the final pass on Standard or above if the deliverable can't carry any visible mark.
3. Don't write exclusive or one of a kind into the client contract; Runway's own terms decline to promise either.
4. Match outputFormat to what the client's editor actually opens, before spending credits re-rendering.
5. Keep the source prompt and seed on file. It won't guarantee an identical reroll, but it's the closest thing to a paper trail if a client asks for a revision six weeks later.

One honest note on our own product. Prompt Architects doesn't touch Runway's licensing terms and can't change what Runway's own Terms of Use say; that agreement is between you and Runway, not us. What it does is give you somewhere to save the prompt template above once, fill in the client-specific brackets per project, and reuse it across your next Runway generation instead of rebuilding the same structure from an old chat thread. If you're pulling client prompts together across agency deliverable work more broadly, a saved library is the part that scales past the first client.