AI Video Prompt Generator (Sora, Veo, Runway, Kling)

A good video prompt is not a magic sentence — it is a structured list of cinematography terms assembled in the order the models weigh most. This generator composes your selections deterministically (the same inputs always produce the same string) and then formats them for the specific model you pick. Nothing is generated and no data leaves your browser.

Token order (camera-led). Text-to-video models react far more to how a shot is described than a still-image model does, so the camera movement and shot type lead the sentence, followed by the subject and its motion:

Camera + shot (lead) → Subject → Action / motion → Setting → Lighting → Mood → Style

For example, a slow push-in and a close-upcombine into the opening phrase “Slow push-in close-up of …”, then the action, setting, lighting, mood and style are appended as comma-separated clauses. Empty fields are skipped, so you only ever get the terms you chose.

Negative prompts differ by model. Veo, Kling and Runway accept a separate negative-prompt field, so exclusions are emitted as their own block. Sora and Luma have no negative field — their documentation recommends phrasing what you dowant, so the tool rewrites each exclusion into a positive instruction (“blurry” becomes “sharp, in focus”). Pika reads exclusions as an inline -neg parameter, and its aspect ratio as -ar.

Aspect ratio and clip length are validated.Each model documents a specific set of aspect ratios and clip lengths. The pickers only ever show the supported options, and the assembler double-checks the chosen values against the model's documented set before writing them — an independent re-parse powers the “limits verified” badge. Where a model reads the aspect ratio from its interface rather than the prompt (Sora, Veo, Kling, Luma), the tool surfaces it as a settings note instead of forcing it into the text. Clip length is always a settings note, because none of these models take duration from the prompt string.

• OpenAI — Generating videos on Sora (prompting, aspect ratios, clip lengths)
• Google — Veo video generation guide (negativePrompt, aspect ratios, durations)
• Runway — Gen-3 Alpha prompting guide (camera-led structure, clip lengths)
• Kuaishou — Kling AI (negative prompt field, aspect ratios, 5s/10s clips)
• Luma — Dream Machine (prompting, clip extension)
• Pika — creation parameters (-ar, -neg)

The per-model conventions, supported aspect ratios and documented clip lengths on this page were last cross-checked against the vendor documentation above on 2026-07-10. They are reviewed whenever a model ships a new version or changes its supported options. Results still vary by model and seed — prompts follow documented conventions, they do not guarantee a specific output.