One balance covers all four tools and every model on the roster, so choosing per shot costs a decision rather than a subscription. Three consequences: you can mix tools freely, failed renders cost nothing, and the real constraint becomes your monthly allowance rather than which tool you're in. Credits reset monthly and don't roll over, which makes the month the planning unit.
The obvious benefit is not paying for four things. The less obvious one is that it removes a decision you'd otherwise make badly — which tool to use, decided on billing rather than on fit.
What does one balance actually remove?
The billing consideration from a creative decision.
On separate subscriptions, switching tools mid-project has a cost attached, so people don't. They use the tool they're paying for, on shots it isn't right for, because the alternative means another line item.
One balance removes that entirely. Creative Studio, Talking Actors, Make a Custom Video and URL → Ad all draw from the same place, and so does every model on the roster. So the question becomes which tool suits the shot, which is the question you wanted to ask.
The same applies per model. Switching from one model to another for a single shot costs a decision rather than a subscription.
A shared balance means tool choice is a creative decision again rather than a billing one.
What are the drivers, per tool?
Four tools, and the drivers differ enough to matter when you're planning.
| Tool | What drives the cost | What's free |
|---|---|---|
| Creative Studio | Clip length and resolution tier, per model. The cost shows before you render | Reference attachments, prompt length, and which model you pick |
| Talking Actors | Script length, plus whether the actor is multi-look. All multi-look creators share one rate | No duration setting exists — the script is the length, so there's nothing separate to pay for |
| Make a Custom Video | Length, resolution tier, and whether you select a real human actor | Prompt length, custom layer phrases, format, hook and setting selections, inline tagging |
| URL → Ad | Length, quality, and actor selection | There's no prompt field, so there's nothing to pay for writing |
Source: Hexcoded product documentation and render costs page, checked September 2026. Credit rates change with model versions.
The pattern across all four: scale costs money and specificity doesn't. Length and resolution drive the price everywhere. Prompt detail, references and layer choices are free everywhere. The only other lever is whether a real human creator is on screen, which adds a small flat fee.
Nothing you write costs anything. Length and resolution cost everything. Plan around that.
What does that mean for a shoot?
One pattern, and it's the same in every tool.
Iterate short and low. Generate the keeper properly. Work out the framing, the blocking and the read at the shortest length and the lowest resolution tier that lets you judge it, then generate the final at what you actually need.
That's not a trick. It's the only workflow the cost structure rewards, and it's available because the cost shows before you render so you're never guessing.
The second habit follows from prompt detail being free: there's no reason to compress a brief. A short prompt saves nothing and costs you precision.
What does a failed render cost?
Nothing, and this is more consequential than it sounds.
Failed renders return credits automatically — you only pay when a clip comes out. So a generation that errors isn't a loss.
What that changes is your willingness to attempt difficult things. A model at the edge of what it handles, an unusual reference combination, a long take on a complex scene — all of those become cheaper to try than they'd be on a platform where a failure still bills.
Where this falls short. A failed render is free. A render that succeeds and isn't good enough isn't. There's no satisfaction refund, so the cost of a bad result is the same as the cost of a good one — which is exactly why iterating at low cost matters.
What's the real constraint, then?
Your monthly allowance, and it doesn't carry forward.
Plan credits reset monthly and don't roll over. On cancellation you keep access and remaining plan credits until the period ends, after which unused ones expire.
So the month is the planning unit, not the project. A series shot across six weeks spans two allowances whether or not that suits the schedule, and credits unused in month one don't help in month two.
Top-up credits behave differently — they're separate from the plan allowance and valid for twelve months from issue. Which makes them the mechanism for a burst of work that exceeds a month rather than a general buffer.
The month is the planning unit. A project that spans two months spans two allowances, and the first one doesn't carry.
What else does the plan gate?
Three things beyond the credit count, and they're easy to miss when comparing tiers.
Resolution. Capped lower on entry tiers, with 4K above them. So a plan can limit what you can deliver rather than just how much.
Model access. Tiered — the entry plan carries a core set rather than the full roster.
Parallel generation. How many generations can run at once, scaling from two on the entry plan up to six. On a shoot day that's a throughput limit rather than a cost one, and it's the least discussed of the three.
That third one matters for planning specifically. Six parallel generations against two changes how a session runs, independently of how many credits you hold.
How should you actually plan?
Five things, in the order they affect the budget.
Check what your plan gates, not just what it includes
Resolution ceiling, model access, parallel generation. A plan can be adequate on credits and inadequate on any of those three.
Decide the delivery resolution before you start
It's one of only two drivers that scale in every tool, and it's plan-gated on top. Deciding it late means either regenerating or renegotiating.
Iterate at the shortest length and lowest tier you can judge
The cost shows before you render, so this is measurable rather than approximate. Then generate the keeper properly.
Write long prompts
Prompt detail, references and layer selections are free in every tool. There is no economic reason to compress a brief, and compressing one costs precision.
Plan by month, and use top-ups for bursts
Plan credits reset monthly and don't roll over. Top-ups are valid twelve months, which makes them the tool for work that exceeds one allowance.
- One balance across four tools and every model. Switching costs a decision, not a subscription
- Scale costs money and specificity doesn't. Length and resolution drive the price everywhere
- Prompt detail, references and layer selections are free in every tool. Don't compress a brief
- Iterate short and low, then generate the keeper. It's the only pattern the cost structure rewards
- Failed renders return credits automatically. A render that succeeds and disappoints doesn't
- Plan credits reset monthly and don't roll over. The month is the planning unit, not the project
- Top-ups are separate and valid twelve months. That's the mechanism for a burst, not a buffer
- Check parallel generation, not just credits. Two concurrent against six changes how a session runs
Yes. Creative Studio, Talking Actors, Make a Custom Video and URL → Ad all draw from the same balance, and so does every model on the roster. So switching tools or models mid-project costs a decision rather than a subscription.
Length and resolution, in every tool. Plus whether a real human creator is on screen, which adds a small flat fee. Prompt length, references, layer selections and model choice are all free — so specificity costs nothing and scale costs everything.
No. Failed renders return credits automatically — you only pay when a clip comes out. What isn't refunded is a render that succeeds and disappoints, since there's no satisfaction refund, which is why iterating cheaply matters.
Plan credits don't — they reset monthly. On cancellation you keep access and remaining plan credits until the period ends, after which unused ones expire. Top-up credits are separate and valid for twelve months from issue.
Three things beyond the credit count. Resolution, which is capped lower on entry tiers with 4K above them. Model access, which is tiered. And parallel generation, scaling from two concurrent generations up to six — a throughput limit rather than a cost one.
No. Prompt length, custom layer phrases, references and layer selections are free in every tool. Compressing a brief saves nothing and costs you precision, which usually means more re-rolls rather than fewer.
One balance, every tool
Four tools and 30+ models drawing from the same credits, with the cost shown before you render and failed renders returned automatically. Switching for one shot costs a decision rather than a subscription.
See the plansMore on tools, features and how the platform fits together in Product.