# Pricing AI video services as a freelancer — a working method > Work out your credit cost per deliverable, price the outcome instead of the render, write scope that survives revisions, and handle client workspace access. Published 2026-09-07 · eroq.ai — canonical: https://eroq.ai/blog/pricing-ai-video-services-as-a-freelancer The first client who asks "so what do you charge for one of these?" usually gets a bad answer, because the freelancer has just watched a 15-second ad render for less than a sandwich and cannot unsee it. That number is real, and it is not your price. Here is how to compute it properly, what to do with it, and how to write scope so the third revision round does not eat the job. ## Step one — cost one deliverable exactly Price your most common deliverable end to end, including the parts you forget. Take a 15-second vertical ad — three shots of five seconds, one product, one voice-over line. - **Product stills** to use as references, six at 10 credits each — **60 credits** - **Draft pass**, each shot once at 3 seconds on Seedance 1.0 Lite, 36 credits each — **3 × 36 = 108** - **Keeper pass**, two takes of each shot at 5 seconds on [Kling 2.5 Turbo](/models/kling-2-5-turbo) at 1080p, 120 credits each — **6 × 120 = 720** - **Voice-over**, a 220-character line on Voice One at 3 credits per 100 characters — 3 blocks — **9 credits** **Total 897 credits.** On a Creator plan, where 6,000 credits cost $49, that is $0.00817 a credit, so the deliverable costs **$7.33**. At the entry pack rate it is $8.97. Now a heavier one — a 30-second brand piece, six shots, finished on the flagship engine: - Ten stills — **100 credits** - Six 3-second drafts on Lite — **216** - Six finals at 5 seconds on Seedance 2.5 at 180 credits, plus a second take on the three hard shots — **6 × 180 + 3 × 180 = 1,620** - A 450-character voice-over on Voice One, 5 blocks — **15** **Total 1,951 credits, about $15.93** at the Creator rate. Two numbers computed from published per-engine prices rather than estimated. The method behind them — marginal seconds, take multipliers, the cost of a finished minute — is in [how much AI video costs](/blog/how-much-does-ai-video-cost). ## Step two — stop pricing the render Your cost of goods on that first deliverable is $7.33. Nobody is buying 897 credits. The client is buying an ad that runs, arriving Thursday, in the format their media buyer asked for, after someone decided which shots tell the story. Cost-plus pricing fails here in a way it does not in physical work, because the material input has collapsed to nearly nothing while the judgment has not. Mark up $7.33 by even 400% and you are charging $37 for a day's thinking. Price the outcome — the finished ad, in scope, on a date — the way you would if you had shot it on a camera. The useful role for the cost number is as a **ratio test**. On a healthy project, render credits are a small single-digit percentage of the invoice. If they are 30% of what you charge, you are reselling compute at a thin margin, and some spreadsheet will beat you at it. Market rates vary too much by country, sector and client size for anyone to publish an honest one, so do not take a number from an article — including this one. Use what you charged for comparable video work, and let the cost line tell you whether the margin is real. ## What the cost number is actually for - **Quoting variants.** When the client asks for a 1:1 version alongside the 9:16, you know a re-render of three shots is 720 credits and the real cost is your time in the edit. - **Absorbing risk.** Budget takes generously. Four instead of two across the 15-second ad is 12 × 120 = 1,440 credits instead of 720 — a difference of about $5.88, which is not a risk worth negotiating. - **Knowing your plan.** Four of those ads a month is 4 × 897 = 3,588 credits, inside Creator's 6,000 for $49. Add three brand pieces and you are at 3 × 1,951 + 3,588 = 9,441 credits, which needs Studio at $149 for 20,000. Credits roll over forever — check the ladder on the [pricing page](/pricing). ## What goes in scope The render cost cannot hurt you. Your hours can. So scope what consumes hours, specifically enough that a disagreement is a reading exercise. - **Concepts.** How many distinct creative directions are included — usually one, with a named price for a second. - **Shots.** The number per deliverable. Six shots is a different job from three, and should not be discovered at the storyboard stage. - **Takes.** A take is a re-render of the same shot with the same intent. Included takes are your quality budget and they are cheap, so be generous — two or three per shot, stated. - **Revision rounds.** A revision is a note on a delivered cut, and two rounds is normal. State what is *not* one — a new script, a new product, a different concept — because that is the boundary that gets stretched. - **Formats and length.** Which aspect ratios, how many cutdowns, and the finished duration. The engines cover six ratios, so format is a scope decision rather than a technical one, and clip length is capped by plan — 20 seconds on Creator, 30 on Studio and above. - **Files.** What you hand over, in what codec, and how long you keep the project recipes so a re-render is possible in six months. ## When a client wants access to your workspace This is the clause people discover too late, so read it twice. **Every individual plan holds exactly two seats.** Hobby, Creator and Studio are all two — paying more buys credits, speed and longer clips, never more people. So when a client asks for a login, they take your only spare seat, the one your editor or partner was using. Your options, in order of how often they are right: 1. **Don't.** Deliver files. Most clients asking for access want to see work in progress, which a shared folder solves. 2. **Publish for review.** A render can be published so a link is enough, then unpublished. 3. **Give them the Viewer role.** Viewer reads the library and usage and cannot spend a credit — the right rung for a client who needs to watch the queue. It still takes one of your two seats. 4. **Move to a team plan** if client access is part of what you sell. Team is $249 for 25 seats and 36,000 credits, Agency $499 for 50 seats and 80,000, on one shared wallet. If you do seat a client, set their limits on the Members tab at [/dashboard/workspace](/dashboard/workspace) — each member can get a share of your rate limits and a monthly credit ceiling, so an enthusiastic client cannot drain the wallet before your deadline. The role ladder is in [workspace roles and permissions](/blog/workspace-roles-and-permissions), the shared-wallet side in [managing credits across a team](/blog/managing-ai-credits-across-a-team). Start by costing your own most common deliverable — open [the video studio](/studio/video), render the three shots, read the ledger. Agency-scale versions of this workflow live in [agencies](/use-cases/agencies). ## FAQ ### How much does it cost me to produce one AI video ad? A 15-second three-shot ad with drafts, two takes per shot, product stills and a voice-over line comes to 897 credits — about $7.33 on a Creator plan. A 30-second six-shot piece finished on the flagship engine is 1,951 credits, about $15.93. Compute your own from the per-engine prices. ### Should I charge clients per credit? No. Credits are your cost of goods, not a unit the client values, and passing them through invites a negotiation about compute instead of about work. Price the finished deliverable and use the credit figure privately as a margin check. ### Can I give a client access to my eroq workspace? You can, but every individual plan holds exactly two seats, so a client seat is your collaborator's seat. Use the Viewer role if they only need to watch, and move to Team or Agency if seating clients is a regular part of how you work. ### How many revisions should I include? Two rounds of notes on a delivered cut is a normal default, with takes handled separately and generously since re-renders are cheap. Write down what counts as a revision versus a new concept — that is the boundary clients push on.