Important API update: OpenAI says the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API was discontinued on September 24, 2026. So this template can help you plan a video scene or adapt your prompt for another tool, but you can’t use it to generate a new video through the Sora API now. (
help.openai.com)
Copyable Sora 2 prompt template
Use this as a creative brief, not an official prompt format or a guarantee of what a model will produce. The fields below are an editorial template based on OpenAI’s archived Sora 2 prompting guidance, which says prompts can guide the shot while leaving room for interpretation. (developers.openai.com)
[Subject] [does a specific action] in [setting].
[Shot type and camera position]: show [important visible details].
[Motion and pace]: [how the subject, objects, or camera move].
[Lighting and mood]: [light quality, time of day, and tone].
[Optional continuity anchor]: keep [distinctive visual detail] consistent.
[Optional audio]: [dialogue, ambience, or sound effects, if relevant].
What to put in each field
- Subject and action: Say who or what is in the scene and what happens. Prefer a concrete action over a general mood alone.
- Setting: Give the location and a few details that help establish it.
- Shot and visible details: Name the framing—such as a wide shot or close-up—and say what the viewer should see. OpenAI’s guide describes prompting as briefing a cinematographer and advises that details can help guide the result, while some openness can invite variation. (
developers.openai.com) - Motion and pace: Describe what moves and how quickly. Add camera movement only if it matters to the scene.
- Lighting and mood: Choose a few cues that fit the intended tone rather than piling on conflicting descriptions.
- Continuity anchor and audio: Optional. Repeat a distinctive item or appearance detail when continuity matters; describe audio only when it is important to the scene.
These are useful prompt-writing suggestions, not required Sora fields. The Sora documentation is archived and explicitly warns that it may refer to outdated models or APIs. (developers.openai.com)
Example: fill in the template
A cyclist rides slowly along a quiet coastal road at sunrise.
Wide tracking shot from slightly behind: show the ocean on the left, the road curving ahead, and a yellow rain jacket clearly visible.
The cyclist pedals steadily while the camera follows at an even pace; sea grass shifts in a light breeze.
Soft golden light and a calm, reflective mood.
Continuity anchor: keep the same yellow jacket and black bicycle throughout.
Audio: gentle surf and faint tire noise; no dialogue.
The example keeps one central action and adds framing, visible details, movement, light, and optional audio. It doesn’t ensure that any particular detail will appear exactly as written: OpenAI’s archived guide cautions that a prompt is a creative direction rather than a contract, and that repeated generations can vary. (developers.openai.com)
How to revise a prompt that misses the brief
- Name the mismatch. For example, the shot is too close, the action is unclear, or the jacket is not visible.
- Change the relevant cue. If the jacket matters, make it a clearly visible detail in the shot description. If the framing is wrong, revise the shot type or camera position.
- Keep other wording stable for the comparison. Change one prompt detail at a time so you can tell which revision affected the result. This is a practical testing approach, not a promise of improvement.
- Simplify if the directions compete. Keep the subject, action, and most important visual details; remove cues that pull the scene in different directions.
For a multi-shot idea, describe the shots in order and repeat the visual anchor in each one. For example: “Shot 1: the cyclist starts on the coastal road. Shot 2: the same cyclist and yellow jacket pass a beach access path. Shot 3: the bicycle stops at a lookout.” Treat this as a way to make your brief clearer, not a guarantee of continuity.
Can you use this with the production Sora API?
No. OpenAI lists the Sora API as discontinued on September 24, 2026. The Sora web and app experiences were discontinued earlier, on April 26, 2026. Therefore, there is no current Sora API request path to provide for submitting this prompt. (help.openai.com)
OpenAI’s Sora 2 prompting guide remains online, but it is marked archived and may describe outdated models or APIs. Do not treat its former API examples, parameters, or model identifiers as instructions for a working production integration. (developers.openai.com)
You can still use the template to develop a scene brief or adapt it for a different video-generation product. Before sending it to another service, check that service’s current documentation for its supported model, input format, parameters, and availability; this template does not establish compatibility with other tools.
Sources
What to know about the Sora discontinuation — OpenAI Help Center