Log in Start free

Prompts6 min read

Image to prompt for Veo 3 and Kling: turn a reference into a shot

A still image can become a strong video shot. Learn the two ways to use image to prompt with Veo 3 and Kling, with six Indian examples.

In short: There are two ways to use an image with Veo 3 or Kling. Either turn the image into a full text prompt and generate a fresh shot in the same style, or use the image as the first frame and write a prompt that describes only motion, camera and sound. The second method keeps the look closest to your image.

To use image to prompt for Veo 3 or Kling, first decide whether the image is a style reference or the actual first frame. As a style reference, convert it into a full text prompt (subject, setting, light, colour, lens) and generate a new shot. As a first frame, give the image to the model and write only what should move, how the camera moves and what we hear.

Two methods, two jobs

Method A: image to text promptMethod B: image as first frame
What you give the modelText onlyThe image plus a short prompt
What the prompt describesEverything: subject, place, light, colour, lensOnly motion, camera move and sound
ResultA new shot in a similar styleYour exact image, brought to life
Best forMood boards, shots you cannot photographProduct photos, your own photos, art

Image input support and its options differ between Veo 3 and Kling versions, so check each model's current limits before you plan around Method B.

The motion-only rule for Method B

This is where most people go wrong. When the image is the first frame, the model already sees the subject, the place and the light. If you describe them again, you can push it to redraw the scene and drift away from your image. Write three things only:

  1. What moves: "steam rises", "her dupatta lifts in the wind", "the flames flicker".
  2. How the camera moves: "slow push in", "static", "gentle orbit left".
  3. What we hear (for models that use sound cues, like Veo 3): "temple bells far away", "rain on a tin roof".

Add a short line like "keep the composition and lighting of the image" to anchor it.

6 examples with Indian images

1. Diwali balcony (Method B, Kling)

The image: a balcony at dusk lined with clay diyas and a paper lantern.

Keep the composition and lighting of the image. The diya flames flicker gently, the paper lantern sways slightly in a light breeze, a few sparks from a distant phuljhadi drift across the background. Very slow push in toward the diyas. Calm, warm, festive.

Why it works: small, natural motions that belong in the scene. Nothing in the prompt redescribes the balcony, so the model keeps your image.

2. Masala chai product photo (Method B, Veo 3)

The image: a glass of masala chai on a brass tray with cardamom pods around it.

Keep the composition and lighting of the image. Steam rises slowly from the glass, a single cardamom pod rolls a little on the tray. Static camera. Sound: a soft clink of the glass on brass, a kettle bubbling in the background. Cosy, appetising.

Why it works: steam is the classic motion for a hot drink. The sound line adds warmth, and the static camera keeps the product framing you already chose.

3. Your grandparents' photo (Method B, Kling)

The image: a recent photo of your grandparents sitting on a charpai in their courtyard, shared with their consent.

Keep their faces, expressions and the composition exactly as in the image. They smile softly, the grandfather turns slightly toward the grandmother, a light breeze moves the neem leaves above them. Very slow pull out. Gentle, natural, no exaggerated movement.

Why it works: gentle, believable movement respects real people. Do not make them speak or do things they never did, and only animate family who agree.

4. Ladakh landscape (Method A, Veo 3)

The image: a wide photo of a blue lake with brown mountains under a deep sky. Here you turn it into a full prompt to create a new shot.

Wide aerial shot gliding slowly over a still blue high-altitude lake in Ladakh, bare brown mountains around it, a deep blue sky with a few white clouds. Harsh clear daylight, sharp shadows on the slopes. A single traveller in a red jacket walks along the shore. Sound: strong wind, nothing else. Vast, quiet, cinematic.

Why it works: the colours and light are pulled from the image, but adding a traveller and an aerial move makes a fresh shot with a story point.

5. Kolkata street photo (Method A, Kling)

The image: an old tram on a rainy Kolkata street, yellow taxis, wet tracks. You want a new shot in this mood.

Medium tracking shot alongside an old tram moving slowly down a rainy street in Kolkata, yellow taxis passing, wet tram tracks shining. Overcast grey light, muted greens and yellows, light film grain. The camera keeps pace with the tram. Nostalgic, realistic, no readable signs.

Why it works: the image gave you the palette (muted greens, yellow taxis, grey sky). The prompt turns that palette into movement.

6. Handmade jewellery on a model (Method B, Veo 3)

The image: a studio photo of a model, who has agreed to AI use, wearing oxidised silver jhumkas.

Keep her face, skin tone and the composition exactly as in the image. She turns her head slowly to the right, the jhumkas swing and catch the light. Static camera. Sound: a soft jingle of the earrings. Elegant, realistic, natural skin texture.

Why it works: the head turn exists to show the product moving. Get written permission from any model before you animate their photo.

Turn your image into a prompt on Real Prompt Generator

Upload your reference to the free image to prompt tool and it writes a detailed text prompt you can use for Method A, or trim down to a motion-only prompt for Method B. From there, the Veo 3 prompt generator and Kling prompt generator add camera and timing options. To generate image to video clips with a model picker in one place, see Kritiverse's AI video generator.

Common mistakes

  • Redescribing the image in Method B. It causes drift. Describe motion only.
  • Too much motion. A photo of a still room should not suddenly have five people walking through.
  • Low-quality images. Blurry or tiny images give blurry video. Start with the sharpest version you have.
  • Animating people without consent. Only your own photos, or family and models who said yes.
  • No label. Mark the result "Made with AI", and use YouTube Studio's altered or synthetic content setting and Instagram's AI label where they apply.

Questions

Is image to prompt free?

The basic image to prompt tool on Real Prompt Generator is free. Video generation in Veo 3 or Kling depends on their current plans.

Which is better for image to video, Veo 3 or Kling?

They behave differently, and both change over time. Run the same image and motion prompt on both and keep the one that holds your image better.

Can I animate an old family photo?

Yes, if your family agrees. Keep movement gentle, and never make people say things they never said.

Do I need a separate prompt for each model?

The core idea stays the same. Adjust wording for each model's style and check their current limits.

Keep reading