Kling V2.1 Master

Kling v2.1 I2V Master — kwaivgi/kling-v2.1-i2v-master

Kling v2.1 I2V Master generates short, high-motion video clips from a single reference image plus a motion-focused prompt. Upload an image, describe what moves (subject, camera, environment), and the model animates the scene while keeping the source frame as the visual anchor. Built for stable production use with a ready-to-use REST API, no cold starts, and predictable pricing.

What it’s best at

• Image-to-video generation with strong visual anchoring to the input image
• Cinematic motion: camera moves, parallax, atmospheric effects, subtle facial/body motion
• Prompt-controlled animation with optional negative_prompt to suppress artifacts
• Fast iteration for 5-second clips (and longer durations if enabled)

Pricing

Equivalent unit price: $0.26 per second

Examples

Inputs

• image (required): the reference image used as the first-frame anchor
• prompt (required): describe motion and camera behavior
• negative_prompt (optional): describe what to avoid (blur, distortions, artifacts)

Parameters

• prompt: the motion direction for the clip
• negative_prompt: optional “avoid list” (quality issues, unwanted elements)
• image: the input image (upload or URL)
• guidance_scale: how strongly the motion follows the prompt (lower = more natural drift, higher = stricter prompt following)
• duration: video length in seconds (commonly in 5-second steps)

Prompting guide (I2V)

Write prompts like a director’s brief, prioritizing motion over static description:

• Subject motion: head turn, breathing, hair flutter, hand movement, walking, reading, etc.
• Environment motion: wind in trees, dust, rain, fog, floating particles, light beams
• Camera motion: slow push-in, orbit, handheld micro-shake, tilt up, dolly left, rack focus
• Continuity constraints: keep identity, outfit, and scene layout consistent with the input image

Good pattern: A short description of the scene, then explicit motion cues, then camera movement, then mood/lighting continuity.

Example prompts

• A mysterious woman reading a spellbook in a dark forest. Camera slowly circles her, faint magical lights float around, glowing runes appear, trees subtly twist in the background, cinematic mist and particles, moody low-key lighting, smooth motion, 5 seconds.
• A street portrait at golden hour. Subtle breeze moves hair and clothes, soft lens flare, gentle handheld camera sway, shallow depth of field, natural facial micro-expressions, 5 seconds.
• A product shot on a table. Camera slow push-in, specular highlights glide across the surface, light dust motes in the air, clean studio feel, crisp focus, 5 seconds.

Negative prompt examples

• blur, distort, low quality
• jitter, warping, melted details, extra limbs, duplicate face
• watermark, logo, subtitles, text artifacts, compression blocks

Best practices

• Use a sharp, well-lit reference image; the model can’t “invent” clean details that aren’t there.
• Keep motion instructions compatible with the image (don’t ask for a full outfit change if you only want animation).
• If results look unstable or over-animated, lower guidance_scale and simplify motion.
• If motion is too subtle, add clearer action verbs (turns, steps, lifts, sways) and specify a camera move.