Image-to-video (i2v) generation takes a single still image and produces a video clip where the contents of that image come to life with motion. The input image becomes the first frame (or a key reference frame), and the model generates subsequent frames that maintain visual consistency with the source while adding movement — people walking, hair blowing, water flowing, cameras panning. The challenge is preserving identity and fine details (a specific face, a logo on a shirt, the exact color palette) across frames without drift or morphing artifacts.
Your uploaded image becomes the exact first frame of the video. Kling 3.0 doesn't reinterpret or approximate your image — it uses it pixel-for-pixel as the starting point and generates motion from there. This means your art direction, color grading, and composition carry through unchanged.
Faces, logos, text, and distinctive patterns in the source image remain consistent throughout the generated video. Kling 3.0's temporal attention mechanism cross-references the source image at every frame to prevent identity drift — the face at frame 150 matches the face at frame 1.
Adjust how much motion Kling 3.0 adds to the scene. Low intensity: subtle breathing, gentle wind, slight camera drift. Medium: walking, turning, moderate environment movement. High: running, dynamic camera sweeps, dramatic action. The slider gives you directorial control over the energy level.
Describe what motion you want in the text prompt: "she turns to look over her shoulder", "camera slowly zooms in", "leaves blow from left to right". Kling 3.0 follows these motion instructions while keeping the source image's content intact.
Kling 3.0 automatically detects and matches your input image's aspect ratio — 1:1, 16:9, 9:16, 4:3, 3:4, and more. No need to crop or resize your source image to fit a fixed video format. The output video matches the input dimensions.
Generate videos from 5 to 15 seconds from a single image. For longer sequences, chain multiple generations in Flow, using the last frame of one clip as the first frame of the next to maintain continuity.
Go to Sunra Image to Video and select Kling 3.0 from the model dropdown.
Drag and drop or click to upload the image you want to animate. Use a high-resolution image (at least 1024px on the longest side) for best results. Kling 3.0 accepts JPEG, PNG, and WebP formats.
Describe the motion you want — not the scene itself (the model can see that from your image). Focus on action: "the woman slowly smiles and tilts her head", "camera pulls back to reveal the full landscape", "waves crash against the rocks". Keep it to 1–2 sentences.
Adjust the motion intensity slider (low/medium/high) and choose the video duration (5s, 10s, or 15s). Lower intensity with shorter duration is safer for preserving fine details. Higher intensity with longer duration produces more dramatic results but may show minor drift.
Click Generate and review the result. Check that identity, text, and fine details are preserved throughout. If motion is too subtle, increase intensity and regenerate. If details are drifting, decrease intensity or shorten the duration.
Whether you're a solo creator, an agency, or a brand — every model adapts to how you work.
She slowly looks up from the book and smiles. A gentle breeze moves her hair. Warm afternoon light shifts slightly as a cloud passes.
Model: Kling 3.0 · Duration: 8s · Motion: Medium · Source: portrait photo
Camera slowly orbits 45 degrees around the product. Soft reflections move across the surface. Background stays softly blurred.
Model: Kling 3.0 · Duration: 6s · Motion: Low · Source: product photo on white
Clouds drift slowly across the sky. Water in the lake ripples gently. Camera holds steady. Birds fly across the distant mountains.
Model: Kling 3.0 · Duration: 15s · Motion: Low · Source: landscape photograph
The warrior swings the sword in a wide arc. Sparks fly from the blade. Camera follows the motion with a slight pan right. Cape billows dramatically.
Model: Kling 3.0 · Duration: 5s · Motion: High · Source: digital illustration
Take existing product photos and convert them to short video clips for social ads. A still product shot becomes a 6-second video with a slow zoom and subtle environment movement — no reshooting, no 3D modeling, no After Effects. Batch-produce video variants from your existing photo library.
Artists and illustrators animate their static artwork for social media. A digital painting of a forest gains gentle wind, falling leaves, and shifting light. A character portrait blinks and breathes. Animated posts on Instagram and TikTok get 2–3x more engagement than static images.
Convert real estate photography into walkthrough-style video clips. A wide-angle interior shot becomes a smooth camera pan revealing the full room. Generate 6-second clips for each room and chain them in Flow for a complete property tour from still photos only.
Filmmakers and animators convert storyboard frames into rough animated sequences. Each drawn frame becomes a 5–10 second animated clip showing the camera move and character action described in the storyboard notes. Produces a working animatic in hours instead of weeks.
| Kling 3.0 i2v | Other i2v Models | |
|---|---|---|
| Identity preservation | Pixel-anchored first frame with temporal cross-attention — minimal drift even at 15s | Sora 2: good but can drift on fine details beyond 8s. Veo 3.1: strong but occasionally alters colors. Seedance 2.0: reliable for faces, weaker on text/logos |
| Motion control | Intensity slider + text prompt for motion direction. Motion Brush available for precise paths | Sora 2: text prompt only. Veo 3.1: basic intensity control. Seedance 2.0: text prompt with limited intensity options |
| Max duration | Up to 15 seconds per generation | Sora 2: up to 20s. Veo 3.1: up to 8s. Seedance 2.0: up to 10s |
| Aspect ratio flexibility | Auto-matches input image aspect ratio, any standard ratio supported | Most models support 16:9, 9:16, 1:1 only. Custom ratios may require cropping |
| Audio output | Native audio generation included (ambient sound and dialogue) | Sora 2: no native audio. Veo 3.1: native audio included. Seedance 2.0: music sync but limited dialogue |
Input images of at least 1024x1024 pixels produce noticeably better video quality. Low-resolution sources (below 512px) can result in soft, artifact-heavy output. If your image is small, upscale it first using Sunra's image tools before converting to video.
The model already sees your image — it doesn't need a scene description. Write prompts that describe movement: "she turns left", "camera pushes in", "rain begins to fall". Scene descriptions ("a woman in a red dress in a garden") waste prompt capacity and can cause the model to reinterpret your image.
If your image contains text, logos, or fine patterns that must remain intact, use low motion intensity. High motion increases the chance of detail drift. You can always increase intensity on the next generation if the result is too static.
Kling 3.0 matches the video ratio to your image. If your image is 4:3 but you need 9:16 for TikTok, crop the image first rather than relying on the model to reframe. Intentional cropping gives you control over what's in frame.
Join thousands of creators, agencies, and brands who use Sunra every day.