अपडेट
केस साक्ष्य: मूल ब्रेकडाउन रिपोर्ट पढ़ें · केस फ़्रेम और शॉट समय देखें
प्रकाशित टूल रिपोर्ट पर आधारित। निर्माण वर्गीकरण विश्लेषणात्मक निर्णय हैं; सुझाए चरण मूल प्रोजेक्ट या सत्यापित रीमेक नहीं हैं।
Motion evidence the report recorded
In the music-video breakdown, the same still life — crumpled ink-stained rice paper, an inkstone and a brush — appears twice. Shot #23 is judged AI-generated video (82%): the review notes slight parallax between the paper, the inkstone and the window shadow, and ink texture that drifts between frames. Shot #32 is judged a still image with pan/zoom (77%): paper, inkstone and shadow move as one rigid image. Those observations are exactly the vocabulary motion-control prompts need: what moves, how far, how fast, and what stays still.
Turn a motion note into a prompt clause
The report’s own prompt for shot #23 reads: “Close shot, slightly high angle, ink-stained crumpled rice paper with an inkstone; slow pan right.” Add the stability notes from the evidence and it becomes a set of motion-control clauses: “Close shot, slightly high angle, a crumpled ink-stained sheet of rice paper, an inkstone and a brush on a grey table. The camera pans slowly to the right; the objects stay in place; the window-grid shadow stays constant.” Direction, speed, subject stability, light stability — one sentence each. Vague words like “dynamic camera” give the model nothing to control.
Use a first-frame image for the look
Kling’s image-to-video mode is strong, and the report’s first-frame discipline applies: write one image prompt that fixes subject, setting, light and framing — the report writes one for every generation unit, such as the paper-and-inkstone still life — generate it, then reference it from the video prompts in that setup. Consistency comes from the shared image, not from re-describing the scene each time. A current image model (MiniMax’s H3 image, for example) can produce that first frame — use whichever image model you already trust; the discipline is fixing the frame before any video call.
What we have and have not tested
Tested: the analysis workflow — the report’s per-shot judgments and motion notes are real published output. Not tested: generating with Kling from these prompts. The report does not attribute any shot to Kling, and we make no claim about specific Kling 3.0 features or credit policy — check the current app before committing a plan.
अक्सर पूछे जाने वाले प्रश्न
Is this a Kling 3.0 prompt guide?
The prompt structure here — one shot, motion as clauses, first-frame image for the look — carries across Kling versions. Check the official Kling pages for what the current release supports before generating.
Do Kling prompts need to be long?
No. One shot per prompt, concrete motion, and a reference image beats a paragraph of stacked adjectives.
पुन: प्रयोग योग्य विश्लेषण टेम्पलेट
Shot # / duration Subject + setting (first-frame image ref) Framing / support (fixed, handheld, …) Camera move: direction + speed + distance What stays still (subject, light, background) Audio cue Keep out: cuts, captions
असली रिपोर्ट के साथ अभ्यास
Open the music-video case and read the camera notes shot by shot: lines like “slow pan right” or “slow push-in” are ready-made motion clauses for a Kling prompt. Compare shots #23 and #32 to see a moving camera next to a moved still.
केस खोलें और रिपोर्ट डाउनलोड करें →