Kling 4.0 reference to video on kling4.run: combine photos, short clips and audio, tag each file with an @mention and fix references that get ignored.

Kling 4.0 reference to video builds one clip from your prompt plus the photos, short video clips and audio you attach in the AI video generator. Each file can lend the clip a face, a product, a camera move or a rhythm, and an @mention in your prompt says which file lends what.

How Kling 4.0 reference to video works

No mode switch: one prompt box handles text to video, image to video and reference to video, depending on what you attach.

Click Add reference beside the prompt box. Add files one at a time, in the order you'll mention them.
Write the prompt, inserting a token such as @image1 wherever you name a file. A prompt is required.
In the settings panel, choose aspect ratio, length and resolution. No auto setting copies a reference's shape.
Generate. The cost beside the button (✦ N) depends on length and resolution, not references; finished videos land in My Videos.

Every file in the stack is sent, mentioned or not, so remove what you don't want used. Images guide how a subject looks but aren't sent as a fixed first frame; describe the opening moment in words.

Kling AI calls its multi-reference input Omni Reference; for the model's published specs, see What Is Kling 4.0?. This page covers the generator on kling4.run.

What each reference type is good for

Type	Use it for	Tips
Image	A face and outfit, a product, a place, or a style frame	Sharp, evenly lit, subject in full view; crop out clutter.
Video clip	A camera move, a body movement, an action's timing	One clear move per clip; say whether to take the motion, camera path or both.
Audio	The beat and mood of music, or a voice's tone	A clean recording, trimmed to the part you need; name its role.

How @mentions work

A mention is a token like @image1 that points the model at one upload.

Type @ in the prompt to open a "Mention a reference" list, then click a file or use the arrow keys and Enter.
Click the @ button below the prompt and pick a file; the token lands at the end of your text.
Tokens appear as chips (Image 1, Video 1) but are sent as lowercase text like @image1, so typed or pasted tokens work the same.

Numbering

Each type is numbered separately, in the order you added the files: a photo, a clip and another photo become @image1, @video1 and @image2.

Removing a file

Removing a file shifts later files of that type up: delete Image 1, and Image 2 becomes Image 1. Tokens already in the prompt, including the removed file's, keep their old numbers, and nothing checks them. After a removal, replace the affected tokens with fresh ones from the list.

Assign roles with clear sentences

A token says which file; the words around it say what to take:

Take a feature: "the face and braided hair of the woman in @image1"
Borrow motion: "the camera repeats the slow crane-up from @video1"
Exclude the rest: "only the camera path from @video1, not its people"

Example prompts:

Character and camera move

The glassblower in @image1 shapes glowing glass in a dim workshop. Take only the camera path from @video1, a slow half-orbit to her face.

Product and mood

The steel water bottle in @image1 stands on a mossy rock by a dawn stream as the camera slides in. Match the calm pacing of @audio1.

Style frame and motion clip

In the ink-and-watercolor style of @image1, a paper crane lifts off a desk and glides out the window along the path in @video1.

Character and location

The violinist in @image1 plays on the stone bridge from @image2 as the streetlights flicker on. Sound: a slow waltz, passing footsteps.

Image, clip and audio

The breakdancer in @image1 performs the footwork from @video1 on a sunset rooftop, each freeze landing on the beat of @audio1.

Recipes for common goals

Same character across a series: add the character photo first each time so it stays @image1, and describe the outfit identically.
Product in a new setting: a product photo plus a photo of the place: "the bottle in @image1 on the counter from @image2."
Music-led clip: upload the track trimmed to the section you need, then write "motion follows the beat of @audio1."

File requirements

Type	Formats	Limits
Image	JPG, PNG, WebP, BMP, TIFF, GIF	Up to 30 MB, each side 300–6,000 px, width-to-height ratio 0.4 to 2.5
Video clip	MP4, MOV	Up to 50 MB
Audio	WAV, MP3	Up to 15 MB

The uploader checks each file on selection and names any problem: format, size, dimensions, clip length or number of files. Files upload when you generate.

Best practices

Name what to take. "The hairstyle and glasses from @image1" says far more than a bare token.
One role per file. A clip picked for its camera move shouldn't also supply the lighting and cast.
Match your references. Similar light, color temperature and style blend better than a studio shot beside a dim phone snapshot.
Let the text tell the story. Files set look, motion and sound; the prompt decides what happens, in what order (prompt guide).

Troubleshooting reference to video results

A reference seems ignored. Say what to take from it and remove contradictions, such as "a red coat" when the coat in @image1 is blue. A sharper photo or simpler clip also helps.

The wrong file shows up. Hover over each file in the stack and compare its label with your tokens, especially after a removal.

Styles clash. A watercolor still and a gritty handheld clip pull against each other; pick one file for the look and say so.

The request is blocked. The content filter can flag a prompt or image; edit the prompt or try another image. For failed runs and credits, see Credits and troubleshooting.

Image to video: animate a single photo.
Prompt guide: write the text around your mentions.
Credits and troubleshooting: costs, failed runs and support.
Ready? Open the AI video generator.

Kling 4.0 Reference to Video and @Mentions Guide