20 free, cinematic video prompts originally written for OpenAI's Sora 2 — structured as cinematographer briefs. OpenAI has since discontinued Sora, so these prompts now work best in Runway, Kling 3, Veo, and other AI video generators. Covers nature, fashion, sci-fi, horror, drama, and more.
OpenAI shut down the Sora app and its ChatGPT integration on April 26, 2026, and the Sora API will be discontinued on September 24, 2026. Sora 2 is no longer available to use as a consumer product. The 20 cinematographer-style prompts below still work beautifully in the AI video tools that are available today — Kling 3, Runway, Veo 3.1, and others. See the full list of alternatives below.
The Sora 2 prompt generator on this page provides 20 free, professionally written video prompts originally designed for Sora 2 — OpenAI's former flagship text-to-video model, known for long temporal coherence, physics-accurate simulation, and up to 120-second clips. OpenAI has since discontinued Sora (see the notice above), but the prompt-writing craft transfers directly to every other major AI video model.
Writing effective AI video prompts requires a different approach from image generation. Modern video models respond best to cinematographer-style briefs — descriptions that specify camera movement, subject action, lighting, atmosphere, and shot duration. Vague prompts produce average results; structured, specific prompts unlock a model's full capability.
Every prompt below is structured with a camera instruction, scene setup, subject action, and atmosphere. Copy any prompt and paste it into your AI video tool of choice — Runway, Kling 3, Veo, Hailuo, or LTX — to generate your video.
Write for Sora 2 like a cinematographer briefing a camera operator, not like a painter describing a scene:
Click any prompt to copy — paste into Runway, Kling 3, Veo, or any AI video generator
A slow, low camera push through an empty city intersection at 5 AM. The street is wet from overnight rain. Traffic signals cycle through green, amber, red with no one watching. Steam rises from a grate in the foreground. The sky behind the buildings shifts from deep navy to the first pale strip of gold. No people. No sound implied. Cinematic wide angle, filmic color grade, held for 10 seconds.
A mid-range shot of a narrow mountain waterfall dropping 30 meters into a pool of clear glacial water. The surrounding rock is dark and mossy. Morning side-light catches the mist and creates a partial rainbow in the lower third. Camera is locked off. The water moves fast and white at the top, slows and spreads into the pool below. Documentary quality, real location feel.
Macro slow-motion video of a mechanical watch being lifted from a black velvet surface. A single thin band of cool studio light catches the brushed stainless bezel and the sapphire crystal. The second hand sweeps. The camera orbits the watch from 3 o'clock to 12 o'clock over 8 seconds. Dark negative space background. Luxury brand aesthetic, extremely high detail.
A handheld medium shot of a woman in her 40s standing at an international arrivals gate, holding a small hand-written sign. She scans the crowd. A man with a travel bag appears from the doors and stops when he sees her. They both pause for one beat before moving toward each other. Natural airport lighting, ambient crowd noise implied. Realistic documentary feel, no slow motion.
A smooth, slow tracking shot through a contemporary art museum at opening time. The camera glides along a white gallery corridor at eye level. Large abstract canvases hang at even intervals. Floor-to-ceiling windows on the left cast long morning shadows across the polished concrete floor. No visitors yet. The camera passes through two archways and ends in a central atrium flooded with skylight.
A telephoto tracking shot of a lone arctic wolf moving through deep snow at dusk. The wolf is white against a blue-gray landscape, heading from frame right to frame left. Its breath clouds the cold air. The camera follows at a constant distance, keeping the wolf in the left third of frame. Birch trees blur in the foreground. BBC Wildlife quality, natural light, no dramatic music implied.
An extreme close-up, top-down shot of a single drop of black ink landing in a glass tank of clear water. The camera captures the moment of impact and the bloom that follows: tendrils spreading outward in fractal patterns, pulling through the water in slow motion. The background is pure white. Shot continues for 12 seconds as the ink cloud disperses into a soft grey gradient.
Ultra-slow-motion underwater side-view of an Olympic swimmer launching off the starting block. The camera is positioned at the waterline — half above, half below the surface. The swimmer's body enters the water in a perfect streamline, arms extended. The splash unfolds in white foam above the surface, while the body cuts through the blue below. Competition pool, lane ropes visible, no crowd shown.
A wide tracking shot moving alongside a model walking across a red sand desert at magic hour. The model wears a structured camel coat, walking away from camera into the low sun. The wind moves the coat hem slightly. The shot holds for 8 seconds. The horizon is flat and empty behind the figure. Color grade: warm, filmic, slightly desaturated. Vogue editorial quality.
A slow forward push down the center hallway of an abandoned Victorian house. The wallpaper is peeling in symmetrical strips. At the far end, a closed door with light bleeding underneath. The camera moves toward it without stopping. The floorboards creak without source. A single portrait on the left wall appears to track the camera's movement. Natural grey light from unseen windows. No music, no jump cut.
An orbital flyby shot of a large modular space station. The camera sweeps from bow to stern over 15 seconds. The Earth curves blue and white below the station's silhouette. Solar panels catch full sunlight on one side; the other half is in shadow. The station rotates very slowly. No starships, no action — just the station, the planet, and the silence of low orbit.
A handheld walk-through shot of a busy Southeast Asian night market. The camera moves through the crowd at eye level, turning past lit food stalls — noodles in broth, skewers over charcoal, fruit piled in pyramids. Vendors call out, steam rises from woks. Warm incandescent light from overhead bulbs. The camera pauses at one stall for 3 seconds, then continues. Authentic documentary feel.
A locked-off medium shot of a man in Victorian dress sitting at a large mahogany desk writing by candlelight. The room is dark except for the candle and a dying fireplace behind him. Bookshelves fill the wall. He writes steadily, pausing once to dip the pen in ink. Wind moves the curtain slightly. Warm amber lighting, period-accurate set dressing, 19th-century atmosphere.
A wide sweeping shot through an enchanted forest at twilight. Glowing mushrooms line the path, fireflies drift between moss-covered trees, and a stream runs beside the path catching the blue-violet sky above. A small wooden bridge arcs over the stream. The camera drifts slowly forward, following the path. Soft bokeh around the light sources. Warm, safe, wonder-filled atmosphere. Suitable for all ages.
A wide, locked-off landscape shot of a desert mesa at night during a lightning storm. Three lightning bolts strike in sequence over 8 seconds — left horizon, center, far right — each briefly illuminating the flat desert floor and the silhouette of a lone saguaro cactus in the foreground. Between strikes: total darkness and stars. No rain in shot. Weather documentary quality.
An extreme close-up, side-angle slow-motion shot of warm dark chocolate being poured from a ladle onto a plain white ceramic plate. The chocolate pools in the center and flows outward slowly. The sheen catches studio light from the upper left. Steam is faintly visible. The pour lasts 4 seconds and then the ladle lifts away cleanly. Food film quality, warm tones, no garnish.
A wide shot of a contemporary dancer in a vast, empty white studio. The dancer is in the center of frame, barefoot, wearing a plain black leotard. The sequence begins with stillness — then moves through a 10-second phrase that includes a floor sweep, a turn, an extension, and a final held position. Even flat studio light from above. No music implied, no audience. The camera does not move.
A sweeping drone aerial shot flying along the top edge of white chalk sea cliffs. The camera is positioned just above the cliff edge looking out to sea. Below: dark blue Atlantic waves crashing at the cliff base. Above: overcast sky. The camera travels laterally for 15 seconds, revealing a lighthouse at the end of the headland. No people on the cliffs. Moody, grey palette. BBC Earth quality.
A locked-off wide shot of a rain-soaked urban intersection at 2 AM. Neon signs in red and electric blue reflect in the puddles across the road surface. A single taxi drives through frame from left to right, its headlights sweeping the wet pavement. No pedestrians. Rain falls steadily but gently. The camera holds for 12 seconds after the taxi passes. Cinematic color grade, high contrast.
A medium close-up of a man in his late 70s sitting alone at a corner table in a small European café. He holds a demitasse of espresso in both hands but does not drink. He looks out the window at something beyond frame. Morning light falls across his face from the left. The café is quiet — ambient sounds implied. The camera holds steady for 10 seconds. Naturalistic, no voiceover, no movement.
How Sora 2 compares to the other top-ranked AI video models for different use cases:
| Model | Max Clip Length | Max Resolution | Access | Best For |
|---|---|---|---|---|
| Sora 2 (OpenAI) | 120 seconds | 1080p | Discontinued (Apr 2026) | No longer available — see alternatives below |
| Kling 3 (Kuaishou) | 15 seconds | 4K native | Kling.ai (paid) | Social content, lip-sync, all-around quality |
| HappyHorse (Alibaba) | 10 seconds | 1080p | API / Alibaba Cloud | Physics realism, #1 without-audio leaderboard |
| Runway Gen-4.5 | 60 seconds | 1080p | Runway.ml (paid) | Post-production workflows, professional VFX |
| Veo 3.1 (Google) | 30 seconds | 4K | Google One AI Premium | Native audio, photorealism, Google ecosystem |
Sora 2 offered the longest clip duration of any major AI video model, but OpenAI discontinued it in April 2026. Of the models still available, Runway offers the longest clips (60 seconds) and Kling 3 / Veo 3.1 lead on resolution (4K).
The Sora 2 prompt generator on this page gives you 20 free, professionally written video prompts for OpenAI's Sora 2 model. Each prompt is structured as a cinematographer's brief — specifying camera movement, scene description, subject action, lighting, atmosphere, and duration. Sora 2 performs best when prompts treat it like a camera operator: describe what the camera sees and does, not just what the subject does.
No. OpenAI discontinued Sora as a consumer product — the standalone app and its ChatGPT integration both shut down on April 26, 2026, and the Sora 2 API will be discontinued on September 24, 2026. There is no longer a way to generate videos with Sora 2, even with a ChatGPT Plus or Pro subscription. OpenAI cited high compute costs and a shift toward its core enterprise products. The cinematographer-style prompts on this page still work well in the AI video tools that are available today — Kling 3, Runway, Veo 3.1, Hailuo, and LTX.
Write for Sora 2 like you're briefing a cinematographer, not describing a painting. Include: (1) Camera type and movement — 'slow push forward,' 'locked-off wide shot,' 'handheld tracking shot'; (2) Scene description — location, time of day, weather, environment; (3) Subject and action — who or what is in frame and what they do; (4) Lighting — natural, studio, practical, direction; (5) Atmosphere — the mood and implied sound; (6) Duration anchor — '10 seconds,' 'the shot holds.' Avoid vague aesthetic instructions like 'beautiful' or 'cinematic' without specifics — Sora 2 needs the specifics themselves.
Sora 2 excels at: long temporal coherence across clips up to 120 seconds; realistic physics simulation (water, fire, cloth, smoke); complex multi-element scenes with many moving parts; atmospheric environmental footage (weather, crowds, natural landscapes); character consistency held over a full shot; and smooth professional camera movements (tracking, push, sweep). It is particularly strong for establishing shots, atmospheric sequences, and slow-motion footage where physics accuracy matters.
Sora 2, Kling 3, and HappyHorse are the top three AI video models by quality as of June 2026. Kling 3 leads the overall leaderboard with native 4K, 60fps, 15-second clips, and built-in multilingual lip sync — it is the best all-around model for content creators. HappyHorse is #1 on the without-audio leaderboard and excels at physics realism and stylized motion. Sora 2 offers the longest clip duration (120 seconds), the best temporal coherence for complex scenes, and the most intuitive natural-language prompting. For pure narrative filmmaking and atmospheric sequences, Sora 2 remains the strongest choice; for fast iteration and social content, Kling 3 is more practical.
Yes. The cinematographer-brief format used in these prompts transfers well to other video generation models. Kling 3 responds strongly to camera movement instructions. HappyHorse handles detailed environment and atmosphere descriptions well. Runway Gen-4 and Veo 3.1 also benefit from structured shot descriptions. You may need to adjust clip duration references — Sora 2 supports up to 120 seconds while most other models cap at 10–30 seconds. All other elements (camera type, lighting, subject action, atmosphere) work across platforms.
Kuaishou Kling 3 — native 4K, 60fps, #1 on global AI video leaderboard
Alibaba's #1 AI video model — best physics realism, without-audio leaderboard leader
Google's latest video model — native audio, 4K, free for Google One AI Premium
Runway Gen-4.5 — 60-second clips, professional post-production workflows
Lightricks open-source 22B model — 4K/50fps, runs locally on NVIDIA RTX
MiniMax Hailuo — #1 physics simulation, anime and cinematic style modes
Edit and caption your clips with CapCut, then add a studio-quality AI voiceover with ElevenLabs.
AI video editing with auto-captions, effects, and one-tap templates.
Try CapCutThe most realistic AI voice generation and voice cloning available.
Try ElevenLabsSome links above are affiliate links. If you sign up we may earn a commission at no extra cost to you. It helps keep this site free.