How Magic Hour’s Credit Balance Covers Video, Image and Voice

Chapter 1

The Product Launch That Started With Three Phone Photos

A product launches this afternoon and all that exists for it are three photos taken on a phone last week – no studio shots, no video, no voiceover. That is a familiar starting point for anyone who has to produce marketing content alone, and it usually means opening three different browser tabs: one for image editing, one for video, one for a voice recording tool, each with its own login and its own bill.

Magic Hour collapses that into one login and, more usefully, one shared balance. Whatever gets spent generating a background-swapped product photo, a short video from that photo, or a voiceover to sit underneath it, all draws from the same pool of credits rather than three separate subscriptions charging separately for the same afternoon's work.

That single balance is worth understanding before anything else, because it changes how the site is actually used. A tool is not chosen because it has its own price tag – it is chosen because it does the job, and the balance simply gets smaller. The free tier hands over a small number of starter credits to try this out without a card on file, which is enough to see whether the whole idea holds up before committing to anything.

What is less obvious on the first visit is how differently priced the tools inside that one balance actually are. A still image and a few seconds of video do not cost the same number of credits, and neither does a voice clip of the same length. Knowing that difference – not the tool list, not the interface – is what decides whether a plan lasts the month or runs dry by the second week.

What you can actually do here

The three subscription tiers and the three credit packs are the two places Magic Hour actually asks for a decision. Both are shown here exactly as the pricing page states them, side by side, so the choice can be made once rather than re-read every time.

Choosing a subscription plan

Topping up without a subscription

Chapter 2

One Balance, Every Tool

Every generation on Magic Hour is priced in credits before it runs, and the number is shown up front – nothing is charged silently after the fact. What is not shown up front, and has to be learned by comparing tools, is that video costs meaningfully more per second than an image does, and some video modes cost more than others for the same length of output.

The rule of thumb the site itself gives is to take the seconds of video wanted and multiply by its frame rate – lower frame rates for animation-style output, higher ones for lip sync, text-to-video and image-to-video. A ten-second animated clip and a ten-second talking-head clip are not the same spend, even though both are '10 seconds of video' to a person planning a shoot.

📰 Recommended for You

How Picsart Turns One Photo Into a Production Line

Read More →

What SeaArt AI Lets You Create Before You Ever Pay

Read More →

This matters most at the boundary between a subscription's monthly allowance and needing more. Because the balance is shared across every tool, a month spent mostly on images leaves plenty of room for one ambitious video; a month spent mostly on video-heavy work can drain a plan that would otherwise have looked generous on paper. The allowance is a pool to budget across tool types, not a per-tool quota.

One detail that changes how that budgeting actually feels: unused credits are not lost at the end of a billing cycle. They carry forward indefinitely, so a quiet month is not wasted spend – it becomes a buffer for a busier one later. That single fact is what makes planning ahead of a big production week worthwhile rather than pointless: credits saved in March are still there to spend in June.

Chapter 3

From a Sentence or a Photo to a Finished Clip

The most direct route into video starts from one of two places: a single sentence describing a scene, or a photo that already exists. Text-to-video takes a written prompt – subject, setting, camera movement, lighting, style – and renders it as a short clip without any footage to begin with. Image-to-video takes something that already exists, like a product photo, and animates it: the prompt in this case describes the motion rather than the whole scene.

Both start the same way after that: a duration and an aspect ratio are chosen (vertical for social feeds, square, or widescreen for a display ad), and the render happens without any timeline, keyframe or clip-trimming step in between. This is the part that replaces an editing tool rather than sitting alongside one – there is no assembly step to hand off to something else.

What decides whether the first attempt is usable is almost entirely the prompt's specificity. A vague prompt like 'a person talking' produces a generic result; naming the camera angle, the lighting mood and what the subject is actually doing produces something closer to a directed shot. This is the same discipline a photographer or director already has – it is just typed instead of set up on location.

A generated clip that ends before the story does is a real ceiling, not a flaw to work around quietly: a video tool has a maximum length per generation, and a scene that needs to run longer has to be extended as a second, separate step rather than requested longer the first time. Knowing that before starting a longer sequence avoids discovering it mid-project.

Chapter 4

The Same Face, Every Time You Need It

A recurring problem in any multi-part content series is that an AI tool generates a slightly different version of the same person every time it is asked – a different face shape, a different expression baseline, sometimes a different skin tone. For a one-off image that is invisible. For a five-part video series featuring the same presenter, it breaks the illusion completely.

Magic Hour's answer is to let a person be saved once, as a Character, from between one and twelve reference photos, and then selected by name inside any tool that generates faces – image generation, image editing, image-to-video and text-to-video alike. The photos are combined automatically into a single reference the tools draw from, rather than being re-uploaded and re-described for every new generation.

The practical effect is that a brand spokesperson, a recurring host, or a founder's face can be set up once and then reused across a month of separate pieces of content without the person needing to be re-photographed or the prompt needing to redescribe their appearance every time. One of the twelve reference photos can be marked as the primary image, which is what tools using a single-face reference fall back on.

Style consistency is a separate but related problem, solved by a second feature called Moodboards: a set of reference images uploaded once to capture a colour palette, lighting mood or texture, then selected the same way a Character is. More than one Moodboard can be applied to a single generation at once, which is how a genuinely mixed look – vintage photography with futuristic lighting, for instance – gets built without describing every visual detail in the prompt each time.

Chapter 5

Giving a Face a Voice, and a Voice a Language

Once a face exists – a Character, a talking photo, or a video already shot – lip sync is what makes it appear to say something new. Audio is uploaded (a voice recording, a generated voiceover, or a preset), and the mouth movements in the visual are matched to it automatically, without frame-by-frame adjustment.

The most immediate use of this is translation without reshooting: the same footage, dubbed into a different language, with the mouth movements adjusted to match the new audio rather than staying locked to the original language's shapes. A training video or a product explainer that needs eight language versions can be produced from one original shoot, run through voice generation and lip sync per language, instead of being filmed eight separate times.

The voice generator behind that dubbing step can either read a script in a standard AI voice or clone a specific voice from a sample. Both draw from the same shared credit balance as everything else, priced by length rather than by which mode was used – a detail worth checking before generating a long script rather than after.

The one thing this workflow does not remove is the need for a decent original recording or video to start from – lip sync adjusts mouth movement to new audio, it does not repair a poorly lit or poorly framed original clip. Where the source footage itself is the problem, the fix belongs in the photo and video enhancement tools covered next, not in lip sync.

Chapter 6

Where the Free Tier Actually Stops

The free plan is a genuine way to try the platform, not a crippled demo – no card is required, and a handful of starter credits are given to every account to generate real output rather than a locked preview. Image outputs from the free tier already carry no watermark. Free video and audio outputs may carry one, and paid plans are what remove it.

The boundary that actually matters commercially is usage rights, not the watermark. Content made on a paid plan comes with full commercial usage rights – it can be used in ads, client work or a product listing. Content made on the free plan is licensed for personal, non-commercial use only, which makes the free tier the right place to test whether a workflow works, and the wrong place to finish something that is going out under a business's name.

Ownership is a separate question again: whoever generates an asset retains full ownership and copyright of it regardless of plan, including free accounts. What a paid plan adds is the right to use that asset commercially – the asset itself was always the creator's.

A subscription that runs out of monthly credits mid-cycle does not simply stop working: the options at that point are to wait for the next billing cycle's allowance, upgrade to a higher tier, or buy a one-time credit pack that never expires. Refunds exist too, but narrowly – a subscription purchased within the last seven days, with fewer than 500 credits used, qualifies for a one-time refund; credit pack purchases themselves are final once bought.

Chapter 7

Which Plan Actually Fits the Work You Do

None of the plan tiers are wrong choices in the abstract – each is sized for a different rhythm of work, and the rhythm is what should decide, not the feature list, since every plan carries the same tool set. What differs between them is how much of the shared credit balance arrives each month, how fast exports render, and how many generations can run at once.

Someone posting occasionally – a monthly product shot, a handful of social clips – is well served by the entry tier and its free rollover; there is little reason to pay for a faster queue that is rarely tested by demand. Someone producing weekly, running several projects at once, feels the difference in concurrent generations and priority queueing long before they feel it in raw credit count.

A team producing daily, or an agency serving several clients from one account, is the case the top tier and unlimited concurrent generations were actually built for – and it is also the case where the API becomes relevant, since it is available on every plan but only pays for itself at genuine volume, where generating through code replaces manually operating the interface for each asset.

The credit maths, not the tool list, is what should decide a plan upgrade. A month that runs out of credits by the second week on the entry tier is telling its own story clearly – that is not a plan to defend by using it more carefully, it is a plan that no longer matches the volume of work actually happening.

Questions readers actually ask

Contact / More useful information from RamthaMedia

Official source links:
Magic Hour