A useful Sora 2 prompt reads like a concise brief for a filmmaker: say what happens, where it happens, and what the shot should look and sound like. Treat details such as camera movement, lighting, and pacing as creative direction—not guaranteed controls.
Important API status: As of November 8, 2026, OpenAI says the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API was discontinued on September 24, 2026. You can still write or adapt a Sora 2 prompt, but you can’t submit it to OpenAI’s discontinued Sora API as a production request. (help.openai.com)
A reusable Sora 2 prompt template
Use this as a writing outline, not as required syntax or an API payload. Leave out sections that don’t matter to your scene.
Create a [kind of video or visual style].
Subject and action: [who or what is visible, and the main action].
Setting: [place, time, and a few relevant details].
Shot: [framing and any camera movement].
Visual treatment: [lighting, color, mood, or texture].
Pacing: [how the action develops].
Sound, if wanted: [dialogue, ambience, music, or sound effects].
Avoid: [a specific unwanted element or contradiction].
OpenAI’s archived Sora 2 prompting guide describes a prompt as a creative brief: more detail can give clearer direction, while leaving room open can invite variation. It also cautions that a prompt is not a contract and that repeated generations can differ. (developers.openai.com)
Example prompt
Create a quiet, cinematic product shot of a handmade blue ceramic mug on a kitchen windowsill on a rainy morning.
Subject and action: Steam rises from the mug while raindrops move down the window behind it.
Setting: A lived-in kitchen with soft daylight and a blurred green garden outside.
Shot: Begin in a medium close-up, then make a slow, subtle push toward the mug.
Visual treatment: Muted blue and warm cream colors, soft natural light, gentle reflections on the ceramic glaze.
Pacing: Calm and unhurried; keep the mug as the clear focal point.
Sound, if wanted: Quiet rain against the window and faint room ambience; no dialogue or music.
This is an illustrative prompt, not a reported test. Its sections help make the intended scene legible; they don’t guarantee that a model will follow every direction or reproduce an exact camera move.
How to make a prompt clearer
- Choose one main action. A single, visible event is easier to describe clearly than several unrelated actions competing for attention.
- Anchor the scene. Name the subject, setting, and a few details that matter. Skip decorative details that don’t help communicate the shot.
- Describe the visual intent. Lighting, color, framing, and mood can help explain the look you want. Treat terms such as “slow push-in” as instructions in prose—not as proof of precise camera control.
- Add sound only when it matters. If your chosen generation tool supports audio, identify the sounds and what should cause them. For another tool, check its own current documentation.
- Resolve contradictions. “Bright, sunlit noir” may be ambiguous unless you explain how those ideas should combine. Decide which direction should dominate.
- Keep pacing plausible. Describe a manageable progression rather than packing a long sequence of actions into a short scene. The source guide recommends describing pacing, but its old timing examples should not be treated as current product limits. (
developers.openai.com)
Simplify an overloaded prompt
Overloaded: “A courier races through a crowded market, stops to buy food, gets into a chase, climbs a roof, and escapes by helicopter, all in one continuous shot.”
More focused: “A bicycle courier threads through a crowded market as vendors pull carts aside. Follow from behind at street level; warm late-afternoon light, lively but readable movement.”
The revision narrows the scene to one central action. It is an editing example, not a claim about a tested Sora result.
Prompt text is not the same as API settings
A creative prompt describes the video you want. In the archived March 2026 guide, OpenAI distinguished that prose from attributes it said had to be set as API parameters, including model, size, and duration. That guide is explicitly archived and may refer to outdated models or APIs. Since OpenAI says the Sora API was discontinued on September 24, 2026, do not use its old model names, parameter values, or examples as current production instructions. (developers.openai.com)
There is no current OpenAI Sora API submission path to document here. If you are using a different video-generation product, check that provider’s current documentation for its endpoint, authentication, supported fields, limits, and output retrieval. A Sora-style prompt outline is not a substitute for that provider’s request schema.
What to expect from prompt wording
A prompt is creative direction, not a guarantee. A model may interpret wording differently than you intend, and repeating the same prompt can produce different results. If you use a currently available generation tool, revise one aspect at a time—such as the action, lighting, or framing—so it’s easier to see what changed. OpenAI’s archived guide recommends iteration but does not promise a particular outcome. (developers.openai.com)
Sources
What to know about the Sora discontinuation — OpenAI Help Center