Kling AI: video from a photo or text in 2026

You know the feeling. A reel needs to go out, there's no time to film, the light outside is gray, and all you have ready are a couple of photos from your last shoot. You could hire a videographer and wait a week. You could stitch the photos into a slideshow nobody watches to the end.

Kling closes that gap in one evening. You upload a photo or describe a scene in words, and the AI makes a real clip: a person turns their head and smiles, steam rises over a cup, the camera glides around a desk with your materials on it. With sound and speech, if you need them. These clips look like an expensive shoot, and people watch them to the end in the feed.

I'll walk through Kling so you can make your first clip while you read. Inside: how to sign up and pay, what credits actually cost, step-by-step instructions for video from text and from a photo, a prompt formula and 12 ready prompts for different tasks, working with sound, 10 business scenarios, and a big section on common glitches. Read in order or jump to what you need from the table of contents.

Section 01What Kling is in 2026

Kling is an AI video generator from the Chinese company Kuaishou. It's the same company behind the Kuaishou video platform, China's second-largest short-video service after Douyin. They have huge experience with short vertical video, and it shows: Kling is especially good at exactly what reels need.

The first version came out in summer 2024. The model has been updated several times since, and Kling 3.0 launched in February 2026. Over the summer it gained a fast, cheaper Turbo version and an Omni mode that can edit a finished video from a text command.

What Kling can do as of fall 2026:

• Video from text. You write what should happen in the frame, and the AI films it from scratch.
• Video from a photo. You upload a picture and it comes alive: people move, fabric sways, the camera travels.
• 3 to 15 seconds per generation. After that you can extend the clip a few seconds at a time.
• Sound with the picture. Footsteps, café noise, splashing water, even a character's speech with lips moving in sync.
• Start and end frame. Give it two photos, and Kling makes a smooth transition from one to the other.
• Multi-shot. Several shots in one clip, following your storyboard, like real editing.
• Character references. Upload a photo of a person or an object, and it stays recognizable across different scenes.
• Motion transfer. Take a video with a dance or a gesture, and Kling repeats that movement on your character.
• Images. There's an image generator inside too, handy for making the starting frame of a future video.

Why people who make clips every day love Kling. It keeps the face from the source photo better than most competitors, facial expressions look natural, and hands and fingers break less often than a year ago. Price is another strength: a second of Kling video usually costs less than in Google Veo or Runway, at comparable quality.

It has weak spots too, and it's fair to name them. Scenes where many people move at once sometimes turn into mush. The physics of tricky objects, like juggling or pouring liquid from one container to another, comes out with mistakes. Kling draws text inside the video poorly, so add captions while editing. And the free tier is slow: at peak hours a clip can sit in the queue for dozens of minutes.

Section 02What Kling costs: free and paid plans

Everything in Kling runs on credits. Each clip deducts them from your balance, and the price depends on three things: which model, which resolution, and how many seconds. Sound and speech add to the price. One handy detail: Kling shows how many credits a generation will take right on the Generate button, before you run it. So there are no surprises.

Here's what that means in plain terms. On Kling 3.0 at 1080p, a second of video without sound costs about 8 credits, with sound about 12. So a 10-second clip with sound runs roughly 120 credits. At 720p it's cheaper: about 6 credits per second without sound. The Turbo version is cheaper still, which makes it handy for drafts. The numbers change with every update, so go by the Generate button.

Is the free tier enough? To try it out and see whether it's your tool, yes. 66 credits a day gets you a couple of short clips in basic quality. For steady work it falls short: there's a watermark in the corner, the resolution is weak for ads, and under the service's rules free clips can't be used commercially. If you plan to run a clip in ads or put it on a sales page, get a paid plan.

Section 03Signing up and paying

Getting started takes three things: open the site, create an account, and pay if you need a plan. All three are straightforward.

The site. The official address is klingai.com. Bookmark it so you always land on the real thing.

Signing up. You can sign in with Google, Apple, or a regular email address. With email, you'll get a confirmation code; if it doesn't arrive within a few minutes, check your spam folder.

The app. Kling has apps for iPhone and Android in the App Store and Google Play. The account is shared: clips you make on your phone show up on the website and vice versa.

Paying. There are two simple ways:

• a regular bank card on the website;
• a subscription through the App Store or Google Play, billed to the card or balance already linked to your phone.

Section 04The Kling interface in two minutes

The interface is simple once you know where things are. I'll use the button names exactly as they appear, so you can find them easily.

• Home – a feed of other people's clips. Useful for inspiration: on many of them you can open the prompt and see how the author described the scene.
• AI Video – the main section. Inside are the tabs Text to Video and Image to Video, plus access to frames, references, and extend.
• AI Images – the image generator. You need it to make a good-looking starting frame when you don't have a photo of your own.
• Assets – your history: every clip and image you've made. This is where you download finished work.
• Credit balance – top right. Click it to see how much is left and when it expires.

The generation window always has the same settings. Model (pick the latest, Kling 3.0, or Turbo for drafts). Quality mode. Clip length in seconds. Aspect ratio: 9:16 for reels and stories, 16:9 for YouTube and your website, 1:1 for the feed. A sound toggle. And the prompt field, where you describe the scene. Below it there's sometimes an extra Negative prompt field – that's where you list what should not appear in the frame.

Section 05Your first video from text, step by step

Let's start with the simplest case: a clip made entirely from a description. A task almost anyone can use: an atmospheric opener for a reel about your morning routine or workday.

Why it worked. "A cozy home office in the morning" sets the place and time. The description of the woman is short, with no extra detail. The actions follow one sequence and fit into 5 seconds: opened, sipped, looked, smiled. Steam over the cup is a small detail that makes the frame feel alive. Camera and light set the mood, and "realistic" keeps it from going cartoonish.

A common beginner mistake is cramming a whole movie into 5 seconds: "she wakes up, walks to the kitchen, makes coffee, sits down to work, calls a client." Kling will try to fit it all in and give you a frantic montage. One or two actions per 5 seconds, three or four per 10 – that's a comfortable pace.

Section 06Video from a photo: bring a picture to life

For business, this mode is more useful than text. You take a real photo of yourself, and the clip stays you: your face, your clothes, your office. No swap for a generated stranger.

The same trick works with objects. A photo of your book or workbook on a desk: "the camera slowly orbits the book, the pages flutter slightly in the breeze, warm lamp light." A photo of a dish: "steam rises over the plate, a drop of sauce slowly slides down, the camera smoothly moves closer." A photo of your office: "a sunbeam slowly moves across the wall, the curtain sways." Clips like these make great reel covers and backgrounds for story text.

Section 07The prompt formula and 12 ready prompts

A video prompt differs from an image prompt. In an image you describe a frozen moment; in a video you describe what changes over time. That's why movement is the heart of a Kling prompt. Here's a formula that works almost every time:

Six blocks, usually 2–4 sentences. You can write more, but every extra item is a chance Kling misreads it. For camera movement, keep a short glossary handy: slow push-in (moving closer), pull back (moving away), orbit (circling around), pan left/right (turning the camera), tracking shot (the camera follows the character), static shot (the camera stays put), handheld (a slight shake, like filming by hand).

Below are 12 prompts for common tasks. Copy them and swap in your own details.

Note the "Mood background for text" prompt. The line about empty space at the top of the frame leaves room for a caption you'll add later in editing. Without it, Kling likes to fill the frame with details from edge to edge.

A word on the Negative prompt field. That's where you write what shouldn't be there: "text, watermark, extra fingers, distorted face, blurry." It doesn't always help, but it cuts down on failed clips, especially ones with people.

Section 08Start and end frame, multi-shot, and extend

Once you've got the basic modes down, these tools give you more control over what happens in the clip.

Start and End Frame. You upload two images, and Kling builds the movement between them. It's the easiest way to get a predictable result. Examples: a photo of an empty desk and a photo of the same desk with your course materials laid out – a clip where everything lands in its place. A photo before a renovation and after. A photo of a closed box and an open one with your product inside. The closer the angle of the two frames, the smoother the transition.

Multi-shot. Several shots in one generation, like in editing: a wide shot, a close-up, a detail. You describe each shot on its own line, and Kling assembles them into one clip while keeping the character and setting. Here's a sample breakdown for 10 seconds:

Multi-shot saves time on stitching and makes the clip look like real footage. It also brings more errors: the more shots, the higher the chance the character changes in one of them. Three shots per 10 seconds is a comfortable limit.

Extend. Like the clip and want it to keep going? Click Extend under the finished clip and describe what happens next. Kling adds a few more seconds while keeping the scene. You can stretch a story to a minute or longer this way, but each extension builds up a little distortion: by the third or fourth step, hands and small details break more often. For reels, one extension is usually enough.

Elements (references). You upload photos of a person, an object, or a place, and Kling uses them as "actors" in a new scene. Handy when you need a series of clips with the same character: your brand mascot, for example, or your logo mug in different settings.

Motion Control. You upload a short video with a movement – a "look over here" gesture or a dance step – plus a photo of a character. Kling repeats the movement with your character. Useful for trending reels where everyone copies the same gesture.

Editing a finished video (Omni). You upload a video and write what to change: "make it evening instead of day," "replace the background with the sea," "remove the person on the left." It doesn't work perfectly, but for small fixes it saves a reshoot.

Section 09Sound, speech, and lip sync

Since version 3.0, Kling generates sound together with the picture. Flip the sound toggle, and the clip gets the noise of waves, café chatter, footsteps, a pen scratching on paper. The sound follows the action in the frame, and the clip instantly feels like it was shot on a camera.

Speech is even more impressive. If your prompt says the character speaks and gives the line in quotes, Kling voices it and moves the lips in sync. Example:

What to know about sound:

• Sound costs more: roughly one and a half times the clip's price. Keep it off for drafts.
• English speech comes out cleanest. Kling speaks other languages too, but the accent and stress can be off. For lines in other languages, it's safer to generate the clip without speech and record your own voice or use a voice clone in ElevenLabs.
• Short lines work better than long ones. One sentence per 5 seconds, two per 10.
• If you need a talking avatar of yourself for several minutes, look at dedicated talking-avatar services like HeyGen. Kling's strength is short, lively scenes.

Skip generating music in Kling: for reels, take a track from the platform's own library. That way you avoid getting blocked over audio, and you ride the trend.

Section 10Let Claude write your Kling prompts

Writing prompts, keeping the formula in your head, and remembering the camera glossary gets tiring. It's easier to hand the job to another AI. Here's what I do: I describe the task in plain words to Claude, and it writes ready-to-use Kling prompts.

If you don't have Claude yet, sign up at claude.ai. For a detailed walkthrough of the first steps, see Claude from scratch: a complete guide for non-techies.

Then you copy the prompts into Kling one at a time. If a clip comes out wrong, go back to Claude: "In the clip from prompt 2 the fog is too thick and you can't see anything – rewrite it so the trees show through." Claude will adjust the wording more precisely than you would on your own.

Two more tricks that save time:

• A reel storyboard. Give Claude the text of your reel and ask it to split it into 5-second shots with a Kling prompt for each. You get a ready shot list for a shoot you never have to film.
• A prompt from a photo. Attach your photo in Claude and ask: "Write a Kling Image to Video prompt so the person in the photo smiles and nods, with the face and clothes unchanged." Claude sees what's in the photo and takes the angle and light into account.

For more ready-made wording for other tasks, check out the AI prompt library, and for turning one idea into a month of reel scripts, see AI for reels: 30 scripts in one evening.

Section 11Where Kling clips pay off: 10 scenarios

A pretty clip for its own sake gets old fast. Here's where Kling clips actually bring views and leads.

The most underrated scenario on the list is the ad creative. Ad creatives burn out in 2–3 weeks, and every refresh used to mean a new shoot. With Kling you make four versions of one clip in an evening: different light, different camera, different first frame. Launch them all, and after a couple of days keep the one with the cheapest lead.

The second strongest is the reel hook. The first seconds decide whether people watch to the end. Live movement at the start – you turning to the viewer, or the camera sharply pushing in on an object – holds attention better than a static image with text. For how to package a whole reel, see social media prompts.

Section 12How not to burn your credits

Kling credits disappear faster than you'd think. A failed clip costs as much as a good one. Here are the rules that save me more than half the budget:

1. Drafts on Turbo at 720p. While you're hunting for the right wording, run cheap tests. Once you've found it, run the final at 1080p on the main model.
2. 5 seconds first. If the 5 seconds work, extend. If not, you've lost half as much as with a 10-second clip.
3. Sound last. Turn sound on only for the final version; on tests it's a wasted expense.
4. The starting frame as an image. Generating an image costs less than a video. First get the perfect first frame in AI Images or another AI, then animate it in Image to Video.
5. Change one thing at a time. If you rewrite the whole prompt, you can't tell what worked. Changed the camera – look. Changed the light – look.
6. Log in every day. You can't save up the free 66 credits, but you can spend them daily. Over a month that's almost two paid Standard bundles.
7. Save the prompts that work. Start a note called "Kling works." In a month you'll have your own library, and clips in a consistent style will start coming out on the first try.

Section 13When something goes wrong

Below are common glitches and what to do about them. Almost everything is fixed by editing the prompt, no technical skills needed.

Section 14Kling vs other AI video tools

Kling isn't the only strong generator. Here's how it compares with the tools people most often weigh it against, as of fall 2026:

How to choose in ten seconds. You need people in frame and lifelike faces for reasonable money – Kling. You need a cinematic clip with complex lighting and the budget allows – Veo. You need fine-grained editing tools – Runway. Many people keep two tools: Kling for the main work and a second one for tasks where Kling is weaker. I covered what happened to Sora and what replaced it in Sora shut down: where to make AI video in 2026, and a broader overview of tools for different tasks is in best AI tools 2026.

Section 15Pre-publish checklist

Before you post a Kling clip, run through this list:

• The format matches the platform: 9:16 for reels and stories, 16:9 for YouTube and your website.
• The face in the last second matches the first.
• Fingers, eyes, and teeth look normal. Watch the clip paused frame by frame.
• There's no garbled text inside the frame. Captions are added in editing.
• No watermark if the clip goes into an ad or onto a sales page.
• For commercial use, the clip was made on a paid plan.
• Only you, or people who gave consent, appear in the photo.
• The sound doesn't clash with the music from the platform's library.
• The first 2 seconds hook: movement or a cut right away, no slow build-up.
• The prompt that worked is saved to your note.

RecapThe whole path in a minute

Start today with one clip: take your best photo, animate it with the prompt from Section 6, and put it at the start of your next reel. Watch what happens to your completion rate. It'll most likely be the cheapest "shoot" of your life.