Video Studio — Overview

Generate AI videos from a prompt or a starting picture — the ways in, the choices, and what a clip costs.

What this page covers

The studio is one screen, and its manual is two pages. This page covers what you meet before you press the button: the ways in, the engine, the lengths and what they cost, the reference pictures and the shape of the clip.

Video Studio — Your Clips and Refused Clips covers what happens after: the Gallery, downloading, the invisible mark, and what to do when a clip fails or is refused.

Shot List — Make a Movie Shot by Shot is about the third chip, Shot list, where you plan a movie shot by shot, each one starting on the last frame of the one before.

What is Video Studio?

The Video Studio generates short AI videos from text descriptions. Pick a model, an aspect ratio, a duration and a style, then generate.

You can also start from a picture instead of a prompt: give it an image and it animates that.

Generated videos are kept in your gallery, where you can play, download or delete them.

The two ways in

The form — model, style, ratio, duration and your prompt, all on one screen. This is the fast route once you know what you want.
The wizard — walks you through the choices a step at a time and writes the prompt with you. Worth using the first few times. It has its own model picker, and everything you set in it comes back to the form: the prompt, the style, the length, the ratio, the model, and whether you are filming from a description or from an image.

Your Kit sits under the wizard's description box: tap a saved person, brand, place or style to put its @tag in, tap again to take it out. Later steps keep the tags in a thin In this video row with a ✕ on each and an Edit prompt link. The people, brands and places you tagged are sent to the studio as pictures, the same as attaching them on the form.

Finishing the wizard makes the clip. Its last step is headed Review & Generate, but there is no second press waiting for you on the form afterwards — pressing Generate Video in the wizard fills the form and sends the job in one go, so the credits go at that moment. Nothing asks you again, and that last step lists the mode, model, style, camera, length and ratio but no price. If you want to see what a clip will cost before you spend, close the wizard and set it up on the form instead, where the cost card sits above the button.

The wizard has one control the form does not

The wizard asks for Camera Movement — pan, zoom, orbit and so on — and your choice is written into the prompt as, for example, "orbit camera movement", so the shot actually moves. The Style you pick on the same step is written in too, as "cinematic style" and so on. If you change the prompt by hand in the review box, your text is sent instead.

That makes the wizard worth a visit even if you normally use the form. Leaving it on static adds nothing, which is the right choice when the movement should come from what is happening in the scene rather than from the camera.

If you prefer the form, you lose nothing: write the movement into your description in the same words.

A clip is priced per second — read the cost card before you press

Every clip is priced per second. The cost card above the generate button spells out the sum: the engine's credits per second, times the seconds, equals This clip will cost. The rate belongs to the engine you picked in step 01 and to the resolution the clip is made at.

The Length buttons show the lengths that engine can really make. Changing the length never changes the engine, so on screen a longer clip always costs more than a shorter one on the same engine. To pay less for the same length, pick a cheaper engine in step 01.

Only the length you have currently selected shows a price on its button — the other buttons show the seconds and nothing else until you click them.

There is also an "Exactly N seconds" box that takes any number. It does not invent a length the model cannot make: it tells you underneath what would actually be produced, and names another model if one lands nearer to what you asked for. On Veo 3.1 that box is hidden and the control says only these lengths on this model — but all three lengths, 4, 6 and 8 seconds, can be clicked, and each has its own price. On the same model a longer clip costs more, so 8 seconds is the dearest.

The total is in the cost summary above the generate button. There is no confirmation step — the price sits beside the button, and pressing the button spends the credits. Read the number before you press, because nothing will ask you again. This article deliberately does not repeat the numbers, so it can never go stale.

From a description, from an image, or a shot list

Step 02 What to film has three chips, and the model list changes with the one you pick, because not every model can do all three.

From a description — writes the whole scene from your description.
From an image — takes a picture you upload and animates it. The prompt then describes the movement you want rather than the contents.
Shot list — swaps the form for a plan of your movie: one card per shot, each with a first frame, a few words and a length, made in a chain. See Shot List — Make a Movie Shot by Shot.

The model picker, and "Auto"

The model is step 01 Engine, at the top of the form above the prompt, because it decides which modes, lengths and shapes the rest of the form offers. The first entry in the list is Auto. Auto is the sensible starting point — and it is VyroAI's choice, not the provider's: pick it and the form quietly fills in the tool's own default model before the job is sent. Pick a named model when you want a specific one — the durations on offer change with it. Veo 3.1 makes 4, 6 or 8 seconds, each with its own price. The buttons are the authority: if only 8 seconds is offered, that is the only length the video service can make for you right now.

The badge beside a model name

Open the model list and a model can carry a small coloured pill that answers one question: does this model burn a visible "AI generated" mark into the picture itself? Some video providers do, on some plans, and it is not something VyroAI can switch off — so the picker tells you rather than letting you find out on a finished clip.

There are two:

No visible AI mark (green) — a test clip was made on this model and came back with no mark drawn on it.
Adds a visible AI mark (red) — it does mark its output. Expect the mark in the picture.

A model nobody has tested yet carries no pill at all. Today that is every model, so you will not see either of these pills yet. The badge only appears once a real clip has been generated on that model and looked at by a person, because promising a clean picture we have not verified would be worse than saying nothing. So no pill means unknown, in both directions — not marked, and not clean.

Auto has no pill either, because until the job is sent there is no model to judge.

What every named model does carry is a pill about sound: speaks (green, with the longest line it takes, for example 15 s speech · speaks) means it can speak a saved character's line in her voice, own sound means it makes its own audio and cannot use her voice, and silent means it makes none. Auto has none of these either.

Reference images — several, each with a role

The Image to video side is no longer one uploaded picture. You get reference image slots. The form draws up to four of them, but the number actually sent is set by the model you picked, and today every model takes exactly one — so slot 1 is the picture that reaches the model. Once one picture is in, adding more is locked: clicking the drop zone, the URL box and both Use a saved character and Use a saved brand buttons stop working (a file you drag onto the greyed drop zone is still added), and the form says "This model accepts 1 reference". Drag several files in at once, or pick several in the file window, and the extra ones do land — greyed out, under "Only the first 1 will be sent" — and nothing is deleted while you decide which one to keep.

Each slot takes one image and a role: Background / scene, Persona / character, Product, Style reference, Logo / brand mark, Prop or Other. The slot then offers a small token chip — @background, @persona, @product and so on — which you drop straight into your prompt to say where that image belongs: "@persona walks through @background holding @product". Arrows on each slot reorder them, and the form echoes your prompt back with the tokens highlighted so you can see they matched.

"What the model receives" — the box that checks your tokens

Once at least one reference image is in a slot, a small box appears under the reference slots, with From an image picked, headed What the model receives. It reads your prompt back to you and colours every @token and every image 1 style reference in it.

There are only two colours and they mean exactly what they look like. A token drawn in the accent colour matched a slot and will be understood. A token drawn in red and struck through matched nothing — a role you have not used, or a slot number past the one image the model actually takes — and the model will simply see that word as text.

With the prompt still empty it says "Write a prompt — the references are attached but unnamed." That is not an error. Your pictures are attached and will be sent; you just have not told the model which is which yet.

The locked "says" row, when a character has a line

Put a saved character in a slot and write her line into the prompt, straight after her name in double quotes — @nova says "we ship in an hour" — and, if that character already has a saved voice, a second row appears at the bottom of that same box behind a padlock: <name> says: "your line". There is no Say this line box any more; it was removed on 7 September 2026.

It is locked on purpose. The line has exactly one home — your prompt — and this row is the copy that is really going out, so the two can never drift apart. Click it and nothing happens; edit the prompt. (On a model that can take her voice, the note under the row still says Built from "Say this line" — edit it there. That box is gone; edit the prompt.)

The sentence under it is the honest part, and it changes with the model you picked:

the model can take her voice — it names her as the voice reference, so lips and voice match;
the model cannot take her voice — a silent model, and also Veo 3.1, which makes its own sound — the row is drawn struck through, and the note says the line is not sent to the model; it is recorded separately and mixed onto the finished clip. The words will be right and her voice will be right, but the lips will not match.

So the strike-through here is not a mistake either. It is the box telling you the line reaches your viewer by a different road than the picture does.

Where it is going — the shape of the clip

Step 04 Where it is going asks what the clip is for instead of asking for a ratio. Three buttons — YouTube (16:9), Shorts / TikTok (9:16) and Feed post (1:1) — set the shape for you, and a shape the model you picked cannot make is not offered at all. Under them, Exact ratio lists every shape the model has, for anything the three buttons do not cover.

On some models the shape changes the price, because it decides the resolution the clip is made at. Read the total again after you pick one. The Length buttons sit in this same step — see the warning above.

Styles

Under Advanced Options there is a Style: Realistic, Cinematic, Anime, Abstract or 3D Render. It is stored with the video so your gallery stays sorted, but it is not sent to the model today — the look comes from your prompt. If you want a cinematic result, write "cinematic" into the description rather than relying on the picker.

The gear next to Wizard — Video Settings

In the header of the form, beside Wizard, is a gear. It opens Video Settings: three dropdowns — Duration, Aspect Ratio and Style — and a switch, Remember as my defaults. The dropdowns change the same things as the form; the switch is the reason to open it. With it on, what you pick here is what the studio starts with next time, so you are not resetting the ratio on every visit. This is where the old profile-page video defaults went.

Read the price warning above before making a length your default: the length you save decides part of the price of every clip.

The credit cost sits in the summary card above the button and updates as you change the engine, the length or the shape. That card is your last look at the price — pressing Generate spends it, with nothing in between. If the number surprises you, pick a shorter length or a cheaper engine; the prompt does not change the price.

Where to next

After you press Generate: Your Clips & Refused Clips — Getting the clip back, and what to do when one is refused.
You already have a picture: Shot List & Movies — Starting from a picture and chaining shots into a movie.
Making more than one clip: Saved Characters — Keeping the same face in every clip.

1 of 5 in Video Studio · Last checked 23 September 2026 · Previous: Bot Will Not Come Online