AI Video Prompt Generator

Plan clear shots, camera motion, dialogue, and sound with our video prompt optimizer. PromptWizz writes prompts and settings suggestions; it does not create or test a video.

Guidance checked 2026-10-20. Editorial guidance.

Editorial duration and aspect suggestions, not confirmed vendor limits.

0 / 10,000

Optimization depth

Intermediate and Complex are available on Starter and Pro.

Text, image, and video share one allowance.

Pick your model

Gemini Omni Flash

Organize the scene around framing, subject, action, context, and style. Specify a single continuous shot when uninterrupted action is intended. Use concise timing cues where useful. Put quoted dialogue, SFX, and ambience in the audio block.

Kling 4.0

Open with the overall look: visual style, setting, and framing. Then write a time-stamped sequence of events (for example 0–4s, 4–9s) that fits the suggested duration. Repeat each character's identifying details, such as clothing, hair, and props, in every segment where they appear, and keep subject names consistent. Put quoted dialogue attributed to the named speaker, sound effects, and ambience in the audio prompt, not the shot prompt; PromptWizz appends it as the closing Audio block. This structure is an editorial organization aid, not required API syntax.

Kling 3.0

Use Scene, Characters, Action, Camera, and Audio sections. For multiple shots, keep subject labels consistent. Anchor dialogue to the speaking subject and action. This ordering is an editorial organization aid, not a required API syntax.

Runway Gen-4.5 — text to video

Describe both what is visible and how it moves, in natural language rather than keyword lists. Cover camera framing and movement, the subject, its action, and the setting, then supporting details such as lighting and style. With no starting image, establish the whole scene. Keep one clear action per beat, and use approximate timestamps only for sequencing, giving each action a realistic duration. Treat requested sound as a separate post-production plan.

Runway Gen-4.5 — image to video

Describe camera movement, subject movement, environmental movement, and pacing. State what remains still when relevant. Preserve the starting image’s appearance rather than redescribing it. Treat requested sound as a separate post-production plan.

Generic video

Describe an ordered shot plan with concrete action, camera behavior, pacing, continuity, and optional audio intent. Keep settings separate and identify duration as a suggestion.

Build a video prompt one shot at a time

Start with who or what stays in view, then describe a physical action that can fit the clip. Order the beats and label shots only when you want a cut. Keep clothing, objects, and location consistent unless the story requires a change. A single continuous shot is often enough.

Camera movement, subject movement, and pacing

Name the principal camera behavior for each shot, such as a slow push-in or a locked wide frame. Then say how the subject moves and what holds still. A pacing word makes the movement easier to interpret than a stack of camera commands.

Text-to-video versus image-to-video

A text-to-video prompt must establish the whole scene; that applies to Gemini Omni, Kling, Runway Gen-4.5 text to video, and generic video. For Runway Gen-4.5 image to video, provide a first frame in Runway and direct motion from that image. PromptWizz cannot inspect or upload it, so review any suggested action that depends on unseen or off-frame content.

Dialogue, sound effects, and ambience

Quote exact dialogue and use stable speaker names. Keep lines short enough for the clip. The audio block joins the target prompt only for profiles with a verified native-audio workflow. Runway (both modes) and generic video show requested sound as a separate production plan.

Choose a realistic duration and aspect ratio

Omni suggests 3–10 seconds, Runway Gen-4.5 suggests 2–10 in both modes, the Runway-hosted Kling 4.0 surface suggests 3–30, and Kling 3.0 suggests 3–15. The generic 3–10 second range is an editorial starting point, not a vendor limit. Choose portrait or landscape to match where the clip will be viewed, then trim action density to fit.

Prompts differ by video workflow

Gemini Omni Flash

Organize the scene around framing, subject, action, context, and style. Specify a single continuous shot when uninterrupted action is intended. Use concise timing cues where useful. Put quoted dialogue, SFX, and ambience in the audio block.

Checked 2026-10-20: scene, camera, dialogue and continuous-shot guidance. Source.

Omni may add cuts unless the shot is explicitly continuous.

Kling 4.0

Open with the overall look: visual style, setting, and framing. Then write a time-stamped sequence of events (for example 0–4s, 4–9s) that fits the suggested duration. Repeat each character's identifying details, such as clothing, hair, and props, in every segment where they appear, and keep subject names consistent. Put quoted dialogue attributed to the named speaker, sound effects, and ambience in the audio prompt, not the shot prompt; PromptWizz appends it as the closing Audio block. This structure is an editorial organization aid, not required API syntax.

Checked 2026-11-03: settings, time-stamped event sequence, repeated identity cues, audio block. Source.

Limits refer to the verified Runway-hosted kling_4 surface; Kling's own app may offer different durations, resolutions, or controls.

Kling 4.0 on Runway always generates audio; there is no audio on/off setting.

PromptWizz cannot see reference images or videos you attach in the destination tool; review anything that depends on them.

Kling 3.0

Use Scene, Characters, Action, Camera, and Audio sections. For multiple shots, keep subject labels consistent. Anchor dialogue to the speaking subject and action. This ordering is an editorial organization aid, not a required API syntax.

Checked 2026-10-20: editorial section ordering. Source.

Limits refer to the verified Runway-hosted surface, not every Kling interface.

At most six suggested shots for that hosted multishot workflow.

Runway Gen-4.5 — text to video

Describe both what is visible and how it moves, in natural language rather than keyword lists. Cover camera framing and movement, the subject, its action, and the setting, then supporting details such as lighting and style. With no starting image, establish the whole scene. Keep one clear action per beat, and use approximate timestamps only for sequencing, giving each action a realistic duration. Treat requested sound as a separate post-production plan.

Checked 2026-11-03: natural-language visual and motion description, optional framing-subject-action-setting order, approximate timestamps. Source.

The Runway API limits text-to-video prompts to 1,000 characters and offers 16:9 or 9:16 only; the Runway web app may offer other options.

Audio instructions are a post-production plan, not a native audio claim.

Runway Gen-4.5 — image to video

Describe camera movement, subject movement, environmental movement, and pacing. State what remains still when relevant. Preserve the starting image’s appearance rather than redescribing it. Treat requested sound as a separate post-production plan.

Checked 2026-10-20: motion-first direction. Source.

Requires a starting image in Runway. PromptWizz cannot inspect it or verify off-frame motion.

Audio instructions are a post-production plan, not a native audio claim.

Generic video

Describe an ordered shot plan with concrete action, camera behavior, pacing, continuity, and optional audio intent. Keep settings separate and identify duration as a suggestion.

Checked 2026-10-20: editorial, not vendor-confirmed.

Editorial duration and aspect suggestions, not confirmed vendor limits.

Veo prompts and the move to Gemini Omni

The documented Veo 3.1 preview identifiers are scheduled to retire on October 22, 2026. This does not mean every GA Veo service shuts down. Gemini Omni Flash is the active Google choice here, with its own duration and cut behavior. Veo 3.1 preview identifiers shut down on 2026-10-22; this does not establish a shutdown of every GA Veo surface.

Four video prompt examples

Omni continuous shot

Gemini Omni Flash

Shot 1: A single continuous medium shot of a cyclist crossing a wet street at dusk. The camera slowly tracks alongside as reflections move over the pavement; the cyclist remains in frame. Ambient noise: light rain and distant traffic. Suggest 6 seconds, 16:9.

Runway motion from a starting frame

Runway Gen-4.5

With the supplied first frame in Runway: the camera slowly pushes toward the subject while rain begins to move across the visible window. The subject holds still. Suggest 5 seconds; match the frame aspect. Add rain ambience in post-production.

Kling dialogue and two beats

Kling 3.0 on Runway

Shot 1: In a station corridor, Mara looks up from the timetable as the camera holds steady. Shot 2: Mara turns to Eli, maintaining the same coat and location. Mara says, "The last train is leaving." Suggest 8 seconds, 16:9.

Generic silent shot

Generic video

Shot 1: A paper boat drifts slowly across a still pond at morning light. Locked wide camera, no cuts; ripples widen while the shoreline remains still. Silent production plan. Suggest 6 seconds, 16:9.

These are authored shot plans, not generated videos or evidence of model compliance.

What the score means

The video rubric checks stable subjects, ordered action, camera direction, timing, continuity, audio intent, and workflow fit. It scores the source idea and the suggested prompt separately. It does not measure final video quality, speech accuracy, or temporal consistency.

Frequently asked questions

Does this tool create videos?

No. It writes a shot prompt, an optional audio plan, and settings suggestions for you to use in a video tool.

Which video tools can I target?

The active choices are Gemini Omni Flash, Kling 4.0, Kling 3.0, Runway Gen-4.5 text to video, Runway Gen-4.5 image to video, and a generic video profile. Model limits are scoped to the surfaces described here.

What happens to Veo 3.1 on October 22, 2026?

The cited Veo 3.1 preview identifiers retire on that date. That notice does not establish a shutdown of every GA Veo service. Gemini Omni Flash is the active Google profile here.

Can I optimize a Runway image-to-video prompt without uploading an image here?

Yes. You must provide the starting image in Runway. PromptWizz cannot see it, so review any suggested motion that depends on off-frame content.

How do I request one continuous shot?

State that the action happens in a single continuous shot with no scene cut, then describe one clear camera behavior and action sequence.

Can it write dialogue and sound instructions?

Yes. For Omni and the verified Runway-hosted Kling surfaces (4.0 and 3.0), the audio block can be part of the target prompt. For Runway Gen-4.5 and generic video, requested sound is a separate production plan.

How long should my clip be?

Choose a duration the selected workflow supports, then fit the action and dialogue to it. The generic 3–10 second suggestion is editorial, not a vendor limit.

Does it support Sora?

Sora is not an active target. Its app and API were discontinued in 2026.

Do video optimizations use the same quota?

Yes. Text, image, video, and extension optimization share one allowance.

Does the score measure final video quality?

No. It evaluates prompt specificity and workflow compatibility, not temporal consistency or the final render.

More tools and reading

Try the image prompt optimizer, text prompt optimizer, or text prompt checker. Browse templates, plans, and our PromptPerfect comparison.

Background: prompting across models (includes historical DALL-E references) and prompting frameworks for planning. This page carries the active video profiles.

    llms.txt