Photo2ReelCOST
Paying for it

Image to video: how you actually get charged for five seconds.

Four services, the same input photo, the same five-second output, four different bills. Not because one is cheaper, but because they count different things. Once you know which thing is being counted, the rate card stops being a mystery and the retry button stops being free.

The short answer

You are billed on one of four units: a generation, a second of output, a credit, or a subscription. The unit decides what a failure costs you. Under per-generation billing, a bad render is a full charge. Under per-second billing, a short test render is cheap. Under credits, the failure cost is hidden inside a conversion rate. Under a subscription, the failure is free until the cap, and the cap is the real price. Your actual cost is never the advertised unit price — it is the unit price multiplied by how many attempts it took.

One short film frame receiving four distinct unlabelled meter paths
The cost of a short render depends on the billing route around it, not only on the seconds in the exported clip.

Four billing models, one output

The pricing pages all look similar because they all lead with a number. The number is meaningless until you know what it multiplies.

SAME OUTPUT · FOUR UNITS 5 seconds · one source photo · one export PER GENERATION 1 press = 1 charge A failed render costs the same as a good one. Cheapest if you get it right the first time. PER SECOND output length × rate Testing at 2s costs 40% of testing at 5s. Punishes long clips, rewards short tests. CREDITS a hidden exchange rate Cost per clip depends on resolution, length, model. Hardest to compare across services. SUBSCRIPTION included, up to a cap Each render feels free. The cap is the price. Best for steady volume, worst for one-offs. The unit is not on the pricing page. It is in the terms and in what happens when a render fails.
What each model counts. The differences that matter show up on the second attempt, not the first.

Two questions separate these faster than any comparison table:

Does a failed render cost me?Ask support, or run one deliberately impossible prompt and watch the counter. This single answer changes the whole calculation.
Is the unit fixed or does it scale?Per-generation is fixed. Per-second and per-credit scale with resolution and length, so the same service can be cheap at one setting and expensive at another.
A simple price meter connected to repeated test frames and a final chosen frame
A visible rate is only one boundary: retries and the route chosen for testing can change what the finished decision costs.

The rate card is not the cost

Here is the part that no pricing page will tell you, because it is not a number they control: your real cost depends on your hit rate, and your hit rate depends on your input, not on the model's quality ranking.

A service at half the unit price with a one-in-three hit rate costs more than a service at full price with a two-in-three hit rate. That is arithmetic, not opinion:

Real cost per usable clipcost per usable clip = (unit price × attempts per usable clip) attempts per usable clip = 1 ÷ hit rate so: cost per usable clip = unit price ÷ hit rate Example with placeholders you fill in from the current rate card: unit price = A hit rate = B (measured, not guessed) cost = A ÷ B If A = 10 and B = 0.33 → 30 per usable clip If A = 18 and B = 0.66 → 27 per usable clip The cheaper-looking service loses.

Two things follow from that formula, and both are more useful than any ranking.

First: hit rate is a property of your input. A clean, well-lit subject on a plain background against a modest motion request succeeds far more often than a busy frame with several subjects and a large camera move. Same service, same price, very different real cost.

Second: you can measure your own hit rate in one session. Take one photo. Run the same prompt five times at the lowest setting the service allows. Count how many you would actually keep. That fraction is B. It takes fifteen minutes and it beats every review you will read, including this page.

WHY A CHEAPER UNIT CAN COST MORE cost per usable clip = unit price ÷ hit rate 40302010 unit price 10, falling hit rate unit price 18, rising hit rate low hit ratehigh hit rate Shape of the relationship, not a measurement of any specific service. Plug in your own A and B.
The lines cross. Which side of the crossing you are on is decided by your own measured hit rate, not by the vendor.
A wide open frame meeting a visible service boundary line before an export frame
A promise about a free route needs a clear boundary for what is included before it can answer a production question.

Why "free unlimited" never has an answer

People search for unlimited image to video constantly. The reason no page gives a satisfying answer is structural, and it is worth stating plainly rather than pretending a secret service exists.

Generating a video frame requires computation that scales with the number of frames. That cost does not go to zero when a service is free. So an offer of unlimited output has to be bounded somewhere else — otherwise the bill grows without a ceiling while revenue does not. The bounds that actually appear are consistent:

A queueUnlimited in count, limited in time. Your tenth render takes longer than your first, and at peak hours much longer.
Reduced settingsUnlimited at a resolution or duration that is lower than the paid tier's. The cap moved from count to quality.
A reset windowA daily or monthly allowance that refills. "Unlimited" means "unlimited within a period", which is a limit with a clock on it.
A visible markFree output carries a watermark or a branding frame. The price is paid in the file rather than in money.

This is an inference from how the cost structure works, not a claim about any named service's terms. The test is simple: read what happens on your twentieth render in a day, not what the headline says. If the answer is a queue, a downgrade, or a mark, you have found the real cap.

What about the assistants you already pay for

This is the question people actually type, in various forms: can the chat assistant I already subscribe to make a video from a photo. The honest answer has three parts.

Sometimes yes, and it is the cheapest path you have. If video generation is included in a subscription you already hold for other reasons, the marginal cost of a render is zero to you. Under the four models above, that is the subscription model with the cap already paid for.

The cap is monthly and shared. Video generation inside a general assistant draws on the same allowance as everything else you do there, and the allowance resets on a calendar boundary. A busy week of chatting can leave you with no video budget, which is a strange failure mode but a real one.

Control is the trade. General assistants are good at accepting a photo and a sentence. They are weaker on the things that decide whether a clip is usable: exact duration, aspect ratio, seed, and the ability to run the same request again with one variable changed. If you need to compare attempts, a dedicated tool with visible settings will save you more time than it costs.

So the practical order is: try the subscription you already have first, measure its hit rate on your actual photo, and only then pay for a dedicated service. Doing it in the other order is how people end up paying twice.

Four things that raise your hit rate, in order of effect

These are the levers that move B in the formula above. They are ordered by how much they tend to matter, not by how easy they are.

  1. Match the frame before you upload.If your photo is portrait and the output is landscape, something has to be cropped or padded, and that decision is being made by software that does not know what matters in your image. Crop it yourself to the target ratio and the model gets the composition you intended.
  2. Ask for one motion.A push-in, or a pan, or a subject turn. Requests that combine several movements give the model more ways to fail, and each failure is a full charge under per-generation billing.
  3. Reduce what has to stay consistent.Faces, text on packaging, and reflections are the three things that break most often. If the shot does not need them sharp, frame them out rather than hoping they survive.
  4. Test at the shortest length the service allows.The first two seconds tell you whether the motion is right. Paying for five seconds to discover the direction is wrong is the single most common waste.

None of these require a better model. All of them raise the hit rate on the model you already picked, which is the only variable in the formula you control without changing vendors.

When switching actually helps

Switching vendors is worth it in a narrow set of cases, and a waste in the rest. The distinction is whether your failures share a cause.

Switch: the failure is motion understandingIf every attempt produces movement in the wrong direction, or the subject drifts when you asked it to hold, that is a model capability gap. A different model can fix it.
Switch: the failure is a hard limitIf you need a length, ratio, or resolution the service does not offer, no prompt will produce it. Check the limits before you spend another credit.
Do not switch: the failure is your inputBlurry source, contradictory prompt, busy background, or a frame that had to be cropped badly. A new vendor inherits all four.
Do not switch: you have not measuredThree bad renders is not a hit rate. Run five at the same settings before you conclude anything about the service.

Before you top up

BEFORE YOU PAY, THREE QUESTIONS 1 · Does the picture itself have to move, or are you arranging photos you already have? Arranging → a local tool costs nothing and has no cap. Moving → continue to question 2. 2 · Does a subscription you already hold include video generation? Yes → use it first and measure your hit rate. The marginal cost is zero until the monthly cap. 3 · Are your failures the model's, or your input's? Input → fix the frame and the prompt; switching vendors inherits the same problem. Only now compare vendors on unit price ÷ your measured hit rate. Re-run the five-attempt test after every change, including a new vendor.
Three questions, in this order. Most people start at the bottom box and pay for an answer they could have got for free.
  1. Write down the unit.Generation, second, credit, or subscription. If you cannot name it, you cannot compare.
  2. Run the five-attempt test on your own photo.Count what you would keep. That number is your hit rate.
  3. Put both numbers in the formula.Unit price divided by hit rate. Compare services on that figure, not on the headline.
  4. Check what a failed render costs.If it costs a full charge, the shortest test length is not optional.
  5. Check the cap, not the word.Unlimited, generous, and flexible all mean something specific in the terms. Find the sentence that names the number.

Where this fits on this site

This page is about the money. The rest of the site is about the shot: how to describe motion so a generation request is not a gamble, and why a result flickers or drifts.

Plan the request firstThe four decisions that make a request usable before you spend anything. Read the planning guide.
Write the motion lineWhat changes between a vague instruction and one that reads as camera work. Motion prompts.
Fix a result that shimmersWhy generated frames flicker and which settings cause it. Flicker.
Keep a face consistentReference frames, angle, and what breaks identity across shots. Face consistency.

And if what you actually need is to assemble photos you already have — with exact durations, a track, and no upload — two tools on this site do that in the browser, free, with no account: a slideshow maker and a vertical clip maker. No generation, no per-second billing, no cap.

No price figures appear on this page on pur pose. Rates and allowances in this category change month to month, and a number written in an article is out of date before it ranks. The formulas above are the part that does not expire.