Kling 4.0 Reference to Video: Images to Consistent Scenes
Last verified: October 29, 2026
Kling 4.0 reference-to-video turns your images of a cast, a product, a location or a style frame into a single clip up to 30 seconds long. It can frame that clip in ultra-wide 21:9 as well as widescreen, vertical and square. Kling 4.0 generates that full length natively, from a few seconds up to half a minute, so you can stage a long take around the same references instead of planning a string of short clips.
In this mode you show the model what matters instead of describing it. Kling 4.0's reference inputs are built to work together, so characters, actions, shots and audio come out closer to what you intended. In practice, a recurring mascot, presenter or product can come from the same images in every clip of a series, so you don't have to rewrite a paragraph of appearance notes each time.
Everything happens in one form. Add the images, and a voice sample if a character speaks. Write a brief that tells Kling 4.0 what each image is for. Then pick the clip's length, resolution and frame shape and generate. The sections below cover what the model does with those images, how to prompt it, and what to change when a result drifts.
Explore more reference to video
Inputs, Settings and Output at a Glance
What you can upload, set and get back from a single run.
Input
Up to 10 images (JPG, JPEG, PNG or WEBP, up to 10 MB each) + up to 3 audio files (MP3 or WAV, up to 25 MB each) (optional) + a text prompt of up to 5,000 characters
Video Duration
3 to 30 seconds
Resolution
720P or 1080P
Aspect Ratio
Auto, 16:9, 9:16, 1:1, 21:9
Output
1 video per run
Prompting Techniques for Reference Images
Assign Every Image a Role Before Describing the Action
Split Longer Clips Into Timed Shots
Direct the Voice and Delivery of Every Line
How to Make a Video From Reference Images
Six steps from your reference images to a downloaded clip.
Add your reference images in "Add Image"
This upload is required. Give each image one clear subject (a face, a product, a room or a style frame), because a cluttered photo makes it harder to say which part the model should keep. Add the images in the order you plan to mention them, since Kling 4.0 numbers reference assets by upload order.
Add a voice sample in "Add Voice" if a character speaks
This step is optional, so skip it for scenes without dialogue. When a character does talk, upload a clean recording of the voice you want: speech only, with no music bed or noise underneath. A clear sample gives the model the most to work from.
Write the brief in "Prompt"
The box holds up to 5,000 characters. Open with one line per image stating its role, then describe the action, camera and sound. If you only have a rough idea, click "Enhance Prompt" under the box to get fuller suggested prompts. Your text changes only when you pick one, so you can compare versions before committing.
Choose "Video Duration", "Resolution" and "Aspect Ratio"
"Video Duration" runs from 3 to 30 seconds in one-second steps and starts at 5. "Resolution" offers 720P (the default) or 1080P. "Aspect Ratio" offers Auto (the default), 16:9, 9:16, 1:1 and 21:9. Test a short clip at 720P first to see whether your references hold, then render the final at 1080P and full length. If you know where the clip will play, pick the ratio yourself: 9:16 for vertical feeds, 21:9 for a letterboxed cinematic frame.
Click "Generate"
Each run returns one video. Before you click, check that every image the prompt mentions is actually uploaded and that any shot timings in the prompt add up to the duration you chose. A mismatch there is the easiest mistake to make in a long brief.
Download the result
Before downloading, check the clip against your references: faces, outfits, product details and room layout. If one element drifted, tighten the prompt line for that image and run it again. That tells you more than adding extra images at once.
Reference Images or a Text-Only Prompt?
When it pays to start from your own images, and when a written prompt alone is the better start.
| Criterion | Kling 4.0 | Alternatives | Best For |
|---|---|---|---|
| Starting material | This workflow starts from your own images, with Add Image required before a run, and those images define who and what appears in the clip. | Kling 4.0 text-to-video starts from a written prompt, and Kling offers it through the same Omni Reference entry point on its own platform. | If you're still exploring a concept and have no assets yet, start with text-to-video; if you already have a cast, product or location photo, start here. |
| Matching an existing design | Each image can carry a look that the prompt tells the model to preserve, such as a character's facial features, hairstyle and clothing. | In text-to-video the look comes from the written description alone, so how closely it matches depends on how precisely the prompt spells it out. | Use reference images when a face, product or set must match a design that already exists; text-to-video suits original characters whose exact look is still open. |
Common Problems and Fixes
What usually causes a reference-driven clip to go off brief, and what to change before you run it again.
The character's face or outfit drifts away from my reference image
Background or props from one of my reference photos end up in the video
Storyboard panels, sketch lines or notes show up in the clip
Add Image won't accept my picture
My multi-shot prompt won't fit in the Prompt box
Frequently Asked Questions
Which image and audio files can I upload?
Add Image is required and takes 1 to 10 images in JPG, JPEG, PNG or WEBP format, each up to 10 MB and at least 300×300 pixels. Add Voice is optional and takes up to 3 audio files in MP3 or WAV format, each up to 25 MB and no longer than 30 seconds.
What makes one generation cost more than another?
The credit cost of a run is calculated from the options you choose and appears above the Generate button once you're signed in. Video Duration and Resolution are the settings that change it, so check the figure after adjusting either one. Enhance Prompt does not use credits.
Can the characters talk in the video?
Yes. Kling 4.0 generates the dialogue, the on-screen performance and the matching lip movements together, with two-channel stereo sound. Describe the setting, say who is speaking and write the spoken lines into the prompt.
What should I upload in Add Voice?
Upload a recording of the voice a speaking character should have. Kling 4.0 takes voice references as a specified voice for a subject, and Kling suggests using them when a particular vocal identity matters. Background music and sound effects can't be uploaded, so describe ambience in the prompt instead. Kling's own examples ask for subtle background sound that doesn't overpower the dialogue.
Can the dialogue be in a language other than English?
Yes. Write the line in the language you want spoken. Kling's own example scripts a Spanish line and asks for lip movements synced to that language and for natural, standard pronunciation without an English accent. The model also comes with expanded multilingual support compared with earlier Kling versions.
How do I keep a product's design from changing between shots?
Upload a clear image of the product in Add Image. Then spell out its distinguishing details in the Prompt and say they must stay the same in every shot. Kling's guide does exactly this for a red sports car: it describes the taillight layout, the plain paint on the hood, the wheel center caps and the steering-wheel surface, then asks that the car's shape and lighting design stay consistent across all shots.