Guides

    Sora 2 Pro Storyboard: the still is the contract (10s, 15s, 25s, 30cr)

    Sora 2 Pro Storyboard is image-to-video. If the label is wrong, every second is waste.

    Versely Team4 min read

    Sora 2 Pro Storyboard is image-to-video. If the label is wrong, every second is waste.

    Sora 2 Pro Storyboard is OpenAI's storyboard scene generator in the catalog: image-to-video, native audio on, 10 / 15 / 25 second clips, 30 credits. The description is "Sora 2 with advanced storyboard scene generation." That is not a text-to-video scratch pad. The still — the plate you already composed — is the contract the seconds will honor or break.

    Storyboard bills in seconds you cannot unsay

    The duration list is short on purpose: 10s, 15s, 25s. There is no 4-second sketch on this row. There is no 6-second peek. You are buying a storyboard-length take with native audio. Thirty credits is the catalog number. If the opening frame is the wrong product, the wrong face, or the wrong type, you did not "try Sora." You paid for 10, 15, or 25 seconds of the wrong picture talking.

    Storyboard and multi-scene are listed features. They do not rescue a plate. They multiply it. A bad still in scene one is a bad still that now.

    OpenAI's other Sora and GPT Image rows live on the OpenAI roster. This page is only the storyboard generate.

    Native audio on 10, 15, or 25

    audio is true. A silent prompt still gets a soundtrack. Write the sound: room tone, who speaks, what hits the table, whether the line is a whisper. If you needed mute B-roll, this is the wrong row — look at a video model whose audio field is off or empty.

    The catalog does not list a max output resolution on this slug. Do not invent 4K. Do invent the duration before you press generate. 10s is the floor. 25s is the ceiling. 15s is the middle. Pick the one the storyboard actually needs.

    Aspect ratios on the row are 16:9 and 9:16. Frame rates listed: 24fps, 30fps, 60fps. Motion levels: Low, Medium, High. Those knobs matter after the still is right. They do not compensate for a still that is wrong.

    Thirty credits for a plate you already chose

    requires_image is false on the spec — the form may not force an upload — but the category is image-to-video. Treat it that way. Bring a locked frame. Describe motion and sound, not a new scene. Re-describing the picture is how storyboard models invent a second product.

    Run it from image-to-video when the job is "this still, now a scene." The prompting surface for this slug is Sora 2 Pro Storyboard prompting. The longer essay on the mode is the storyboard guide. None of those pages turn the row into text-to-video.

    If you do not have a plate, generate a still first. Then come here. Skipping that step is how 30 credits become a mood film of a product that does not exist.

    Multi-scene is not salvage

    Premium, storyboard, multi-scene — those words in the feature list describe how the take is structured. They are not a quality wand. If the identity in the still is wrong, every scene inherits it. If the label on the bottle is already mush, 25 seconds of native audio will not typeset it.

    The test is simple. Hold the still at full size. Would you print it? If no, do not storyboard it. If yes, write the camera and the sound, pick 10 / 15 / 25, spend the 30 credits once.

    If the still would not print, do not buy 10, 15, or 25 seconds of it talking.

    FAQ

    Is Sora 2 Pro Storyboard text-to-video?

    No. The catalog category is image-to-video. Native audio, 10 / 15 / 25 second durations, 30 credits. If you needed a prompt-only clip with no plate, pick a text-to-video row.

    Does it require an image upload?

    The spec sets requires_image to false, but the job is still image-to-video. Bring a locked still anyway. A row that can run without a file is not a row that should invent the product.

    What happens if I leave the sound out of the prompt?

    You still get native audio. You just did not choose it. Name the voices, the room, and the hits, or you will spend 15 seconds arguing with a soundtrack the model improvised.

    Why not use a cheaper 5-second model to test the still?

    Because this row does not offer 5 seconds. The floor is 10. Test the still on a stills model. When the plate is signed, pay 30 credits for the storyboard length you actually need.