Sora 2 Guide: Video Generation, Characters and Availability

If you search for sora 2, you probably want to know whether it can turn text into video, whether sound is generated along with it, and what sora character actually means. The short answer: Sora 2 is the video and audio generation model officially described by OpenAI. However, as of the review date of 2026-09-21, the Sora website and app have ended, and the Sora API is scheduled to end on 2026-09-24. This article therefore first explains the capabilities it once offered, and then separates the information that is still valid from outdated tutorials.

1. What is Sora 2?

Sora 2 is a model for generating video and audio. The official system card says that, compared with the previous generation, it focuses on physical effects, realism, synchronized audio, controllability and range of styles. In plain terms, you are not just asking for a moving picture; you are describing a short clip scene that includes a subject, an action, camera movement and sound.

Its difference from an ordinary chat model is that the output is not text paragraphs but visuals and sound along a timeline. If your prompt only says "a cat running down the street," the model has to guess the scene, speed, camera and mood on its own. If you add the subject's appearance, the rhythm of the action, camera movement, lighting and ambient sound, the result is usually easier to check and revise.

"Usually" here is a suggestion about working method, not an official guarantee. Generative models can still misunderstand left and right, continuous motion or complex physics, so do not treat every output as a precise animation storyboard.

2. What can Sora 2 do?

According to official material, the core use of Sora 2 is generating video from text or other visual material, optionally with synchronized audio. For a first attempt, you can split tasks into three types:

• Text-to-video: describe a new scene, for example a person waiting for a train at a station in the rain.
• Image-to-video: use an existing picture as a starting point and make the elements in it move. The actual input formats and limits available should be based on the product interface at the time.
• Character or person consistency: use features such as Characters to bring a consented person's appearance and voice into a scene. This is not a feature where you can upload anyone's photo and use it freely.

The most common failure in video generation is putting too many conflicting requirements into one attempt. In initial testing, it is recommended to specify only one subject, one main action and one camera movement, confirm continuity first, and then add background, music or style one item at a time.

If you want to understand common video generation workflows first, you can read the site's AI video tools comparison, and then the image-to-video tutorial to understand the difference between the first frame, reference images and motion descriptions. Those articles describe general methods and do not mean you can still log in to Sora 2.

3. What is sora character?

Sora Characters is the name the official side later used for the character feature; early material also wrote it as cameo. The official description positions it as consent-based control over a person's appearance and voice: the owner of a character can decide who may use it and can revoke that permission, and the owner can view, delete or report videos that include the character.

This design is different from asking the model to imitate a real person. If the material involves a real person, you should first obtain that person's consent and also follow the product's usage rules. The official safety material also notes that content involving real-person likenesses, minors and voices has additional restrictions. Do not treat Characters as a way to get around portrait rights or content safety rules.

Pay attention to the timing: the description of the Characters feature comes from official material from when the Sora product was still running. The Sora website and app ended on 2026-04-26, so as of the review date of this article, it is unverified whether an entry point still exists for ordinary users to create new characters, and unverified whether any third-party service supports Sora Characters.

4. How should beginners write a Sora 2 prompt?

Below is a format that is convenient for repeated revision. It is not a fixed template officially prescribed:

Subject: a mail carrier in a yellow raincoat
Action: walks briskly along an old street, stops at a red mailbox and posts a letter
Scene: evening, light drizzle, wet ground reflecting the street lamps
Camera: start with a wide tracking shot, then slowly push in to the hands
Sound: rain, footsteps, distant traffic; no narration
Rhythm: natural movement; the subject's appearance stays consistent across camera cuts

The point of this format is not to pile on adjectives, but to make each line carry only one kind of information. In the first version, keep five items: subject, action, scene, camera and sound. If problems appear with hands, text signs or multi-person interaction, shorten the scene or change only one variable at a time. For videos intended for brand or commercial use, you should also separately check material licensing, the person's consent, music rights and source labeling in the output.

5. Can Sora 2 still be used as of 2026-09-21?

Do not treat tutorials from 2025 or early 2026 as the current situation. The OpenAI Help Center shows that the Sora website and app ended on 2026-04-26; the same article states that the Sora API is scheduled to end on 2026-09-24. This means that on the review date of 2026-09-21, only a few days remain before the scheduled API end, and any claim that you "can sign up now and use it long-term" should not be stated as fixed without rechecking.

If you created content in Sora before, the official Help Center still provides an export entry: go to sora.chatgpt.com/sunset and select Export. Export timing, data retention periods and whether your account can still operate should be confirmed on your account page and in the latest official notices; this article does not assume the status of any individual account.

6. Which related models does BazaarLink currently have?

Before writing this article, on 2026-09-21, the BazaarLink public model list https://bazaarlink.ai/api/v1/models was queried. The list did not include sora-2, nor any Sora video model ID that could be confirmed from it. Therefore, this article does not provide Sora 2 API integration examples, and it does not claim any other model is Sora 2.

If your need is to first make a static storyboard or concept images, the same public list currently shows these image-related IDs: gemini-2.5-flash-image, gemini-3.1-flash-image and gemini-3-pro-image. These are image model IDs, not video models, and they do not mean they can replace Sora 2. Before formal use, re-query the public list again, because the available models may change.

Official sources checked (review date: 2026-09-21)

• OpenAI: Sora 2 System Card
• OpenAI: Creating with Sora safely
• OpenAI Help Center: What to know about the Sora discontinuation
• Google: Gemini API models (image model endpoint notes)
• BazaarLink public model list

Unverified items: whether Sora Characters still has a creation or management entry after the product ended; the compatibility of Sora 2 with any BazaarLink endpoint; and whether the image models above suit your specific storyboard workflow. This article does not cite prices.

TWD billing · Taiwan invoices · leading AI models · OpenAI-compatible API