Important before you start: As of November 9, 2026, you cannot submit a production request to OpenAI’s Sora API: OpenAI says it discontinued the API on September 24, 2026. Its Sora web and app experiences were discontinued earlier, on April 26, 2026. So this guide can help you draft or adapt a Sora 2-style video prompt, but it cannot provide a currently usable OpenAI Sora API request. (help.openai.com)
A practical structure for a Sora 2 prompt
Treat the prompt as a concise creative brief, not a list of API parameters. One useful drafting structure is:
Subject and action + setting and time + shot and composition + camera movement + visual style and lighting + physical details + sound or dialogue, if wanted
For each part, say what the viewer should see or hear:
- Subject and action: Who or what is on screen, and what do they do?
- Setting and time: Where and when does the scene take place?
- Shot and composition: What is in the frame, and what is prominent?
- Camera movement: Describe the intended movement in ordinary language.
- Visual style and lighting: Name the overall look, palette, and light.
- Physical details: Describe the sequence of contact, movement, and consequence when those details matter.
- Sound or dialogue: Include only the audio elements you want, such as a short line, ambience, or a sound effect.
This is a prompt-writing aid, not an official Sora schema. OpenAI described Sora 2 as able to follow intricate instructions across multiple shots and generate speech, sound effects, and soundscapes. That describes the model’s reported capabilities; it does not guarantee that a particular phrase will produce a particular shot or sound. (openai.com)
Worked example: turn a brief into a prompt
Brief: A courier pauses in a rainy city alley, hears a train, then turns toward the sound. The scene should feel cinematic and tense.
Draft prompt:
At night in a rain-soaked city alley, a courier pauses beneath a neon sign and looks toward the far end of the street. Begin in a medium shot, then slowly push closer as the courier turns after hearing a distant train horn. Reflections shimmer on the wet pavement; cool blue light contrasts with the sign’s red glow. Keep the action restrained and the alley uncluttered. Sound: steady distant rain and one muffled train horn. Tense, cinematic mood.
Why it is structured this way:
- “A courier pauses” establishes the subject and initial action.
- “In a rain-soaked city alley” establishes the location and atmosphere.
- “Begin in a medium shot, then slowly push closer” describes a desired framing and camera move; it is not a hard camera control.
- “Turns after hearing a distant train horn” makes the action sequence explicit.
- The final sentences set the look and sound without adding several competing events.
This is an illustrative prompt, not a tested output or an assurance that Sora—or another system—will render the scene exactly as written. A Sora 2-focused third-party article recommends similar shot-list-style descriptions, but its examples are not official instructions or verified results. (openai.com)
Revise a prompt when the result misses the brief
Compare the result with your intended scene. Name the biggest mismatch, change the wording most relevant to it, and compare the next attempt. This is an editing method, not a guarantee of improvement or repeatable output.
| What seems wrong | A revision to try | What to inspect next |
|---|---|---|
| The action is hard to follow | Describe fewer actions in a clear order: “the ball hits the rim, rebounds, and rolls away” | Does the main action read clearly from beginning to end? |
| The framing feels crowded | Reduce background details and state what should remain prominent | Is the intended subject easy to see? |
| The camera move feels too busy | Remove extra movement cues; state one intended move in plain language | Does the movement distract from the action? |
| Sound competes with dialogue | Shorten the spoken line and reduce the number of simultaneous sound cues | Can the words and key sound be distinguished? |
These are sensible prompt revisions to test, not documented Sora API controls. OpenAI’s Sora 2 launch article used a missed basketball shot and rebound to illustrate more physically grounded behavior compared with earlier video models; it also says the model remained imperfect. (openai.com)
Can you submit a Sora 2 prompt through OpenAI’s production API?
No—not through the discontinued OpenAI Sora API. OpenAI’s Help Center gives September 24, 2026, as the API discontinuation date. Therefore this tutorial cannot responsibly give you a current Sora endpoint, model identifier, authentication example, request payload, or video-retrieval steps. Those would suggest an available route that OpenAI says has ended. (help.openai.com)
If a different provider advertises Sora 2 access, treat that as a separate integration. Before using it in production, check that provider’s current documentation for the model actually offered, API endpoint, authentication, supported inputs and outputs, pricing, and retrieval workflow. Do not assume its availability, parameters, or limits match OpenAI’s former API or Sora app.
Constraints and controls: what not to assume
- Do not treat descriptive wording as a guaranteed control. Camera, lighting, timing, and style phrases communicate intent; this guide does not establish a fixed mapping from a phrase to an exact result.
- Do not carry over old Sora limits. The original Sora research and later consumer-product details are not proof of the former Sora 2 API’s limits. This guide does not claim current Sora 2 values for duration, resolution, or aspect ratio.
- Do not assume a particular API feature exists. A prompt example does not establish support for negative prompts, seeds, image inputs, or any other API field.
- Separate model capability from service availability. OpenAI’s 2025 Sora 2 materials describe video-and-audio generation capabilities, but its current Help Center records that the Sora API and app have since been discontinued. (
help.openai.com)
Quick checklist
Before you use a Sora 2-style prompt with a currently available video tool:
- State the main subject and action.
- Give the setting and the visual mood.
- Describe the desired framing and camera movement in plain language.
- Make important physical actions or cause-and-effect steps clear.
- Add only the sound or dialogue that matters to the scene.
- After generation, identify one mismatch and revise that part of the prompt.
- Confirm the chosen provider’s current API documentation before building a production workflow.
The prompt structure can help organize a video brief, but OpenAI’s Sora API is no longer a production option as of November 9, 2026. (help.openai.com)
Sources
What to know about the Sora discontinuation — OpenAI Help Center