---
title: Video Studio
description: Text and image to video, photo animation, talking photos, video editing, reference clips, upscaling and motion control — modes, models, limits and prices.
icon: Clapperboard
---

Video Studio (`/create-video`) brings every video task onto one page: generate a clip from a description, turn your photo into a video, animate a portrait or make it talk, edit a finished clip, shoot a new one that follows a reference motion, upscale it, and transfer motion from a video onto your character. The form with the model and settings is on the left, the canvas with tips, progress and the result is on the right. The price in credits is always shown on the button before you run anything.

<Callout type="info" title="Video Studio is in beta">
Access to the studio is opening gradually, and the set of tabs and models may change. Each tab has its own address: Text to video is `/create-video`, the other tabs are `/create-video/<tab>`, for example [/create-video/image-to-video](/en/create-video/image-to-video).
</Callout>

## How the screen works

- **Tabs** at the top: Text to video, Image to video, Animate a photo, Edit video, Motion control, Turnkey video. Some tabs have sub-modes — a switch above the form.
- **The form** on the left: model, description, files and settings. At the bottom is the run button with the price and a line explaining it, for example "Price for 5 s of 720p video with sound" or "Price per clip — duration does not change it".
- **The canvas** on the right. Before you run anything, it shows a guide: three steps, three tips and an example of "what to upload → what you get" (Show guide / Hide guide). While a clip is being made, it shows progress in percent; afterwards, a player and a Before / After or Original / Video comparison.
- **My tasks** lists your videos, as well as images and audio, with the statuses Queued, In progress, Ready and Failed, plus a type filter.
- **The form draft** is saved in your browser. If your session expires, sign in again — your description and settings are kept.

## Tabs and modes

<ModeList>
  <Mode icon="Type" title="Text to video" models="Grok Imagine Video 1.5, Veo 3.1 Lite, Kling 3" price={<>Grok Imagine Video 1.5 — 5 credits for 5 s</>} href="#text-to-video">
    Describe a scene and the model shoots the clip. Some models can work from a document, to your own audio track, or across several scenes.
  </Mode>
  <Mode icon="Image" title="Image to video: single image" models="Grok Imagine Video 1.5, Hailuo 2.3 Fast, Kling 3" price={<>Grok Imagine Video 1.5 — 5 credits for 5 s</>} href="#image-to-video">
    Your photo becomes the first frame, the description sets the motion.
  </Mode>
  <Mode icon="ArrowRightLeft" title="Image to video: first and last frame" models="Veo 3.1 Lite, Kling 2.5 Turbo, Veo 3.1 Fast" price={<>Veo 3.1 Lite — 11 credits per clip</>} href="#image-to-video">
    A smooth transition from one frame to another.
  </Mode>
  <Mode icon="Images" title="Image to video: references" models="Veo 3.1 Fast, HappyHorse R2V, Gemini Omni" price={<>Veo 3.1 Fast — 22 credits per clip</>} href="#image-to-video">
    Several images — a character, an object, a setting — are combined into one scene.
  </Mode>
  <Mode icon="Smile" title="Animate a photo: motion" models="Grok Imagine Video 1.5, Hailuo 2.3 Fast, Seedance 1.5 Pro" price={<>Grok Imagine Video 1.5 — 5 credits for 5 s</>} href="#animate-a-photo">
    A smile, a nod, a head turn, a gentle breeze — pick a ready motion or describe your own.
  </Mode>
  <Mode icon="MessageCircle" title="Animate a photo: talking photo" models="Kling Lip Sync" price="from 30 credits" href="#animate-a-photo">
    The portrait speaks your voice recording, with lips moving in sync.
  </Mode>
  <Mode icon="PenLine" title="Edit video: with a prompt" models="HappyHorse Edit, Kling O3 Transform, Gemini Omni" price="by source seconds" href="#edit-video">
    Changes the look of a finished clip — clothes, style, setting — while keeping the motion and the cuts.
  </Mode>
  <Mode icon="Copy" title="Edit video: from a reference" models="Seedance 2.0 Mini, Seedance 2.0 Pro, Kling O3 R2V" price="new video + reference seconds" href="#edit-video">
    Shoots a new video with the motion and camera of the reference, but with your character.
  </Mode>
  <Mode icon="ScanLine" title="Edit video: upscale" models="Clipia 4K Upscale" price={<>10 credits per 5 s at 2K</>} href="#edit-video">
    Enlarges the frame 2× or 4× and sharpens it.
  </Mode>
  <Mode icon="PersonStanding" title="Motion control" models="Kling 3.0 Motion Control, Kling 2.6 Motion Control" price={<>Kling 2.6 Motion Control — 9 credits for 3 s at 720p</>} href="#motion-control">
    Transfers a person's motion from a video onto the character in your photo.
  </Mode>
  <Mode icon="Film" title="Turnkey video" models="the AI director picks them" price="script and estimate cost no credits" href="/en/docs/studios/video-agent">
    Idea → scene-by-scene script → a finished video with voice-over, music and subtitles.
  </Mode>
</ModeList>

"From" prices are calculated for the shortest duration and the lowest quality. Details on every model are in [Video models](/en/docs/models/video).

## Your first clip in three steps

<Steps>
<Step>
### Describe the scene

Open the Text to video tab and write who is in the frame, what they are doing, where it happens and how the camera moves. For example: "A ginger cat sits on a windowsill watching the rain, the camera slowly pushes in, warm evening light".
</Step>

<Step>
### Choose a model and settings

Grok Imagine Video 1.5 is selected by default. Set the duration, quality, aspect ratio and sound if the model supports it. The price on the button updates.
</Step>

<Step>
### Click Generate

A clip is usually ready in 1–3 minutes. Watch it on the canvas, download it, or try another model: each model, and even each run, gives a different result.
</Step>
</Steps>

## Form fields

| Field | What it does | Where it appears |
| --- | --- | --- |
| Scene description | Who is in the frame, what they do, where, how the camera moves. Each model has its own length limit — if the text is too long, the studio shows by how many characters | All modes except upscaling |
| Duration | Seconds from the model's list. Wan 3.0 also offers Auto — the model picks the length | Generation, From a reference |
| Quality | 480p, 720p, 768p, 1080p or 4K — whatever the model supports | Generation, From a reference, Motion control |
| Aspect ratio | Frame proportions. Auto uses the photo's proportions so nothing gets cropped | Text to video and Image to video |
| Sound | "Music and sound effects are generated with the picture." Some models always include sound in the price, others charge extra for it, others have no sound | Text to video and Image to video |
| What to avoid | What must not appear in the frame: "blur, extra fingers, text in frame" | Where the model supports it |

Some models show their own settings under **Advanced settings**:

| Setting | Options | Models |
| --- | --- | --- |
| Speed | Normal or fast: the clip is ready sooner but costs more | Wan 3.0 |
| Web search | On or off. Works only in Text to video, without uploaded files | Seedance 2.0, 2.0 Fast, 2.0 Mini |
| Prompt enhancement | The model adds details to your description. On by default | Wan 2.7 |
| Fixed camera | The camera stays still, only the scene moves | Seedance 1.5 Pro |

## Text to video

The core mode: you describe a scene and get a clip. The default model is Grok Imagine Video 1.5; Veo 3.1 Lite and Kling 3 are also recommended.

Some models can do more:

- **A clip from a document (Wan 3.0).** Upload one PDF, Word, Excel, PowerPoint, Keynote, Pages, Numbers, TXT or MD file — up to 100 MB and 50 pages. The model makes a clip based on the document's content. Such a clip takes longer than usual, and on-screen text may come out distorted.
- **Your own audio track (Wan 2.7).** An MP3 or WAV of up to 15 seconds and 10 MB is added to the clip as its track. It does not change the price. If the recording is longer than the clip, its ending will not make it into the video.
- **Scene breakdown (Kling 3).** No breakdown, Automatic (the model splits the clip into shots itself) or Custom scenes: 2 to 6 scenes, each 1–12 seconds (whole numbers) with up to 500 characters of description, 3–15 seconds in total. The price is based on the total length of the scenes.
- **Characters (Kling 3).** Up to 3 characters who look the same throughout the clip. The name uses Latin letters, digits and underscores, starts with a letter, up to 20 characters; in the description you call a character with `@name`. Each one gets a description of up to 500 characters and 2–4 images from different angles, each at least 300 px. References and characters together — no more than 7.

<Callout type="idea" title="One scene, one action">
Models handle one action per shot best. Build a story of several scenes with Kling 3's Custom scenes or the [Turnkey video](/en/docs/studios/video-agent) tab.
</Callout>

## Image to video

Pick a method above the form:

| Method | What to upload | Notes |
| --- | --- | --- |
| Single image | One photo up to 10 MB, including HEIC from an iPhone | With Wan 2.7 you can upload an **opening clip** instead of a photo: MP4 or MOV, 2–10 seconds, frame side 240–4096 px, up to 100 MB. The model continues it to the chosen length. The duration in the form is the length of the whole video including the opening, and the price is based on it; the result must be at least one second longer than the opening. Wan 2.7 also accepts a recording that drives the motion (MP3 or WAV, up to 15 seconds and 10 MB) |
| First and last frame | Two frames — the start and the end | Not every model supports this: only suitable ones stay in the list. Kling 3's scene breakdown does not work with a frame pair |
| References | Several images: a character, an object, a setting | **Veo 3.1 Fast** — up to 3 images, 16:9 or 9:16 only, 8 seconds. **Seedance** with a single image uses it as the first frame — the references mode needs two or more, and the aspect ratio comes from the frame. **HappyHorse R2V** — up to 9 references |

For image to video, describe only the motion, in one or two sentences. Don't like the motion? Change the description, not the photo. The Auto aspect ratio uses the photo's proportions, so nothing gets cropped.

## Animate a photo

### Motion

1. Upload a photo. A shot with one face or object in the center works best.
2. Click a ready motion — Smile, Blink, Nod, Head turn, Look away and back, Tilt and smile, Gentle breeze, Camera push-in — or describe your own in one phrase.
3. Click Animate photo. In a minute or two, compare Original and Video.

For old and black-and-white photos, turn on **Keepsake clip** — calm motion without speech. All models of the Single image mode work here; Grok Imagine Video 1.5, Hailuo 2.3 Fast and Seedance 1.5 Pro are recommended.

### Talking photo

1. Upload a front-facing portrait with a clearly visible face.
2. Upload a voice recording: MP3, WAV, OGG or AAC, up to 15 seconds and 10 MB. No recording? Generate a voice-over in the [Audio Studio](/en/docs/studios/audio) and come back.
3. Click Make it talk. The clip is as long as the recording, and the price depends on its length — from 30 credits.

The model is Kling Lip Sync, 720p. If the lips do not match the speech, use a sharper photo.

<Callout type="warn" title="Consent and rights">
Animate photos of other people only with their consent, and use only voices and recordings you have the rights to. See the [content rules](/en/docs/product/rules).
</Callout>

## Edit video

| Sub-mode | What to upload | What you get | Limits | How the price is calculated |
| --- | --- | --- | --- | --- |
| With a prompt | A video (you can pick one from My tasks) and a single edit: "make the jacket red", "anime style". Optionally, style images | The same clip with a changed look: motion and cuts are kept, the story does not change, the length equals the source | File up to 50 MB; anything longer than 15 seconds is trimmed. Style images: HappyHorse Edit — up to 9, Kling O3 Transform — up to 4, Gemini Omni — up to 5 (source up to 30 seconds, up to 10 are used) | By source seconds. HappyHorse Edit and Kling O3 Transform have a minimum price even for shorter videos. Gemini Omni charges one price per clip |
| From a reference | A reference clip with the motion and camera you want (a dance, a walk, a fly-around), images of the character, object or setting, and a "What to shoot" description | A **new** video with the reference's motion. It is not an edit or extension of the source | See the table below | For the new video plus the reference seconds. With Kling O3 R2V the result is as long as the reference |
| Upscale | A finished clip from My works, or Use in… → Upscale on a result | The same clip, sharper and larger | Up to 60 seconds; longer clips are trimmed. 2× — up to 1080p–1440p; 4× — up to 4K, only for sources of 720p or lower | By clip length and scale factor |

From a reference limits:

| Model | Reference clips | Reference length | File size | Reference frame | Images |
| --- | --- | --- | --- | --- | --- |
| Seedance 2.0 (Mini, Fast, Pro) | up to 3 | 2–15 s each | up to 50 MB | 480p to 720p | up to 9 |
| Seedance 2.5 | up to 10 | 2–30 s, up to 30 s total | up to 100 MB | — | up to 30 |
| Wan 3.0 R2V | up to 5 | 1–15 s, up to 15 s total; reference and new video together up to 30 s | up to 100 MB | — | up to 10 |
| Kling O3 R2V | 1 | 3–15 s, MP4 or MOV | up to 100 MB | side 700–4553 px, 24–60 fps | at least 1 required |

Some models also accept **audio references** — MP3 or WAV up to 15 MB: Seedance 2.0 — up to 3 recordings of 2–15 seconds, Seedance 2.5 — up to 10 recordings of 2–30 seconds, Wan 3.0 R2V — up to 5 recordings of 1–15 seconds, 15 seconds in total.

Tips:
- Make one clear edit at a time. Do the next one on the new clip.
- For From a reference, use a short clip with clear motion. Seedance 2.0 needs a 480p–720p reference frame: downscale a 1080p phone video first.
- First and last frame cannot be combined with a reference.
- Upscaling sharpens and enlarges the frame, but does not change the story or bring back lost detail.
- You cannot extend a clip with an edit yet. To continue a clip, use Wan 2.7's opening clip in the Image to video tab.

You can also enhance photos and videos on a separate page — see [Enhance and upscale](/en/docs/studios/enhance).

## Motion control

<Steps>
<Step>
### Upload a photo of the character

The head, shoulders and torso should be clearly visible, ideally full-length with nothing blocking the body. The photo must be at least 300 px.
</Step>

<Step>
### Upload a video with the motion

3 to 30 seconds, one person in the frame, up to 20 MB. The resulting clip is as long as the video.
</Step>

<Step>
### Choose the orientation and run

As in the video — for clips up to 30 seconds; As in the photo — for videos up to 10 seconds. Click Transfer motion: the price is charged for every second of the video.
</Step>
</Steps>

The models are Kling 3.0 Motion Control (default) and Kling 2.6 Motion Control (cheaper), at 720p or 1080p. The closer the pose in the photo is to the first pose in the video, the fewer distortions you get.

## Turnkey video

Describe your idea in your own words (3 to 2000 characters): what the video is about, who it is for and where it will be published. Optionally add up to 8 materials — product photos, a logo. Pick 9:16, 16:9 or 1:1 and a length of 15, 20, 30, 45 or 60 seconds, then click Start — script first. The AI director asks a few questions, shows a scene-by-scene script and an estimate before anything runs. The script and the estimate cost no credits; building the video is available on paid plans. Details are on the [Turnkey video](/en/docs/studios/video-agent) page.

## Upload limits

| What you upload | Limit |
| --- | --- |
| Any video file | up to 100 MB — the overall cap; some modes allow less |
| Source for editing with a prompt | up to 50 MB, up to 15 seconds (longer is trimmed) |
| Video for motion control | up to 20 MB, 3–30 seconds |
| Clip to upscale | up to 60 seconds (longer is trimmed) |
| Opening clip (Wan 2.7) | MP4 or MOV, 2–10 seconds, frame side 240–4096 px, up to 100 MB |
| Photo | up to 10 MB, HEIC supported; the exact model limit is shown under the upload field |
| Voice recording for a talking photo | MP3, WAV, OGG or AAC, up to 15 seconds and 10 MB |
| Audio track (Wan 2.7) | MP3 or WAV, up to 15 seconds and 10 MB |
| Audio references | MP3 or WAV, up to 15 MB |
| Document for a clip (Wan 3.0) | up to 100 MB and 50 pages |

## Generation progress and results

- After you run a generation, the request is sent and progress appears in percent. A clip is usually ready in 1–3 minutes.
- You can close the page — the clip is saved in My tasks and then in [My works](/en/docs/product/my-works).
- If a clip takes longer than 30 minutes, the page stops waiting and suggests tracking it in My tasks. The generation is not cancelled.
- The finished clip opens in a player labeled "model · seconds · quality". Next to it are Copy description and Repeat.
- **Download** as MP4 video or the last frame as PNG — handy as the first frame of your next clip.
- **Use in…** passes the clip on: Edit video, Upscale, Motion control, Voice it in Audio Studio, Add to Turnkey video, Publish to feed. If the hand-off fails, download the clip and upload it manually.

Video Studio has no button to cancel a running generation.

## Pricing

<Callout type="info" title="Prices">
Prices on this page come from the service pricing and are refreshed together with the documentation. The exact cost is always shown on the run button before any credits are charged.
</Callout>

The price depends on the model, duration, quality, sound and the length of uploaded files. The line under the button tells you what you are paying for:

- **duration and quality** — most models;
- **per clip** — Veo 3.1 and Gemini Omni edits: duration does not change the price;
- **per second** — motion control;
- **recording length** — talking photo;
- **source seconds** — editing with a prompt, with a minimum price for some models;
- **new video plus reference seconds** — From a reference with Seedance 2.x and Wan 3.0 R2V;
- **length and scale factor** — upscaling;
- **sound surcharge** — for models where sound is optional: "The sound surcharge is already included in the price on the button."

<PriceTable
  caption="Example prices per clip"
  rows={[
    { name: 'Grok Imagine Video 1.5 — 5 s, 480p', credits: 5, note: 'no sound' },
    { name: 'Grok Imagine Video 1.5 — 5 s, 720p', credits: 9, note: 'no sound' },
    { name: 'Seedance 1.5 Pro — 4 s, 480p', credits: 3, note: 'with sound — costs more' },
    { name: 'Veo 3.1 Lite — 4–8 s clip, 720p', credits: 11, note: 'sound included' },
    { name: 'Veo 3.1 Fast — 4–8 s clip, 720p', credits: 22, note: 'sound included' },
    { name: 'Seedance 2.0 Mini — 5 s, 480p', credits: 17, note: 'sound included' },
    { name: 'Kling 3 — 5 s, 720p', credits: 36, note: 'sound costs extra' },
    { name: 'Kling Lip Sync — talking photo, 3 s', credits: 30 },
    { name: 'Kling 3.0 Motion Control — 10 s, 720p', credits: 60 },
    { name: 'Clipia 4K Upscale — 5 s, 2× / 4×', credits: '10–15', note: '2× → 4×' },
  ]}
/>

All models and formulas are in [Video models](/en/docs/models/video).

How the price on the button works:

1. **The price on the button is a server quote valid for 10 minutes.** That exact amount is charged when you run, as long as you have not changed the model or settings. If you change settings or wait longer than 10 minutes, the price is recalculated.
2. **The "from N" label in the model list is a guide.** Only the button shows the exact price. Guests see an estimate; the exact price appears after you sign in.
3. **Auto duration in Wan 3.0** reserves credits as for 30 seconds and charges for the actual clip length — the rest comes back.

The 5 welcome credits you get at sign-up cover one 5-second 480p clip on Grok Imagine Video 1.5. If your balance is too low for a video, the studio suggests choosing a plan or starting with an image — welcome credits cover several images. Plans and packs are in [Credits and plans](/en/docs/product/credits).

## Credit reservation and refunds

- When you run a generation, credits are **reserved**: held but not yet charged. When the clip is ready, the reservation is charged. If the generation fails, the reservation is released in full automatically and your balance updates right away.
- Subscription credits are spent first, then purchased ones.
- Credits are refunded if the generation ends with an error (including when the model refuses under content rules), the clip does not appear within the time limit, the result could not be saved or failed the safety check, or the job got stuck.
- If the description fails the check before the run, no credits are reserved at all.
- My tasks shows the reason for a failed job. Refunds happen on their own — there is no need to contact support.

## Concurrent generations

How many videos, images and audio jobs can be in progress at the same time — jobs that are Queued or In progress count:

| No plan | Basic | Standard | Pro | Ultima |
| --- | --- | --- | --- | --- |
| 2 | 5 | 7 | 10 | 20 |

When the limit is reached, a new generation will not start until one of the current ones finishes. Every model is available on every plan — only your credit balance limits you.

## Troubleshooting

| What you see | What to do | Credits |
| --- | --- | --- |
| "The request was rejected by moderation. Change the description or choose another file: explicit scenes, violence and real celebrities are not generated." | Rewrite the description or replace the file | Not charged or refunded |
| "Not enough credits for this run. Top up your balance — the price is shown on the button." | Top up, or pick a cheaper option: shorter clip, lower quality, no sound | Not charged |
| "Your welcome credits are used up. Top up your balance or choose a plan to continue." | Open [plans](/en/tariffs) | Not charged |
| "Too many requests. Wait a minute and run it again." | Wait a minute | Not charged |
| "This model is unavailable right now. Choose another one." | Choose another model with the same mode | Not charged |
| "Could not start the generation. Please try again." | Run it again | Not charged |
| "The generation failed. Change the description or the model and try again." | Rewrite the description or switch models. If several clips are in progress, wait for one of them to finish | Refunded automatically |
| "No response came from the server, so it is unclear whether the video started…" | **Do not run it again.** Open My tasks: if the video is there, it is already paid for | If the run did not go through, nothing is charged or the credits come back |
| "The clip is taking longer than usual. Track it in “My tasks”." | Wait: the generation continues | — |
| "No connection to the server. Check your internet and try again." or "Your session has expired. Sign in again — your description and settings are saved." | Check your connection or sign in again | — |
| File errors: too large, not a photo, photo too small, video too long | Choose a file within the [upload limits](#upload-limits) | — |
| Upscale: the price has changed and the run did not start | Check the new price on the button and run again | Not charged |
| Upscale: "Processing did not finish. The credits for this run are being returned to your balance." | Try again later | Refunded |

Explicit scenes, violence and real celebrities are not generated; content involving minors is a serious violation. After a safety check is triggered, generation may be temporarily restricted, and for longer after repeat violations. See the [content rules](/en/docs/product/rules).

## Prompt tips

- Describe four things: who is in the frame, what they do, where it happens, how the camera moves.
- One scene, one action. For a story of several scenes, use Kling 3's Custom scenes or Turnkey video.
- For image to video, describe only the motion, in one or two sentences.
- Photo animation works best with one face or object in the center.
- For a talking photo, use a front-facing portrait with a clearly visible face.
- For motion control: a full-length character, one person in the video, and a photo pose close to the first pose in the video.
- For edits: one clear change at a time.
- Didn't work out? Try another model: each model and each run gives a different result.

Video Studio has no camera, lens or director-style presets: describe the camera, lens and shooting style in words in the description — for example, "slow camera push-in, wide-angle lens, handheld shot".

## Model pages

Every video model has its own page at `/create-video/<model>` with a description and examples, and from there you can go straight to generating with that model. Prices and credit rules are the same as in the studio.

## FAQ

<Accordions>
<Accordion title="How long does a clip take?">
Usually 1–3 minutes; clips from documents and long clips take longer. You can close the page — the clip is saved in My tasks.
</Accordion>
<Accordion title="Are the welcome credits enough for a video?">
Yes, for one 5-second 480p clip on Grok Imagine Video 1.5 — that is 5 credits. A talking photo costs from 30 credits, which is more than the welcome credits.
</Accordion>
<Accordion title="Why did the price change?">
The price depends on the model, duration, quality, sound and the length of uploaded files, and the quote on the button is valid for 10 minutes. You are charged exactly what the button shows when you run.
</Accordion>
<Accordion title="The generation failed — are my credits lost?">
No. On failure the reservation is released in full automatically; the reason is shown in My tasks.
</Accordion>
<Accordion title="The connection dropped when I ran it — should I run it again?">
No. Open My tasks first: if the video is there, it is already paid for.
</Accordion>
<Accordion title="What is the difference between With a prompt and From a reference?">
Editing with a prompt changes the look of the source clip while keeping its motion and cuts. From a reference shoots a new video that repeats the reference's motion and camera, with your character.
</Accordion>
<Accordion title="How do I keep the same character across a clip?">
In Kling 3, add Characters — up to 3, with 2–4 images each — and call them with `@name` in the description. Or use the References mode in the Image to video tab.
</Accordion>
<Accordion title="Does the clip have sound?">
It depends on the model. Veo 3.1, Seedance 2.x, MiniMax H3, Wan 3.0 and Gemini Omni generate sound together with the picture. Kling and Seedance 1.5 Pro add it optionally, for a surcharge. Grok Imagine Video 1.5 and HappyHorse have no sound. The With sound filter finds the right models.
</Accordion>
<Accordion title="Can I extend a clip?">
Not with an edit. With Wan 2.7 in the Image to video tab, you can upload a 2–10 second opening clip and the model will continue it to the chosen length.
</Accordion>
</Accordions>