Updated August 7, 2026

How Long Can Sora AI Videos Be?

The current OpenAI API lists 4-, 8-, and 12-second base clips. Longer results can use video extensions, but OpenAI does not publish one universal maximum stitched duration in the API reference.

Short answer: a new Sora API clip currently uses a duration value of 4, 8, or 12 seconds. OpenAI also documents a video-extension endpoint, and the returned video duration represents the stitched total. That makes “How long?” a workflow question as well as a single-generation limit.

Current Sora Video Length Limits

New API clip

Allowed duration values are 4, 8, and 12 seconds in the current create-video reference.

Extended video

The API includes an extension operation. Its duration field reports the stitched total after extension.

Universal maximum

The official API reference does not state one universal maximum stitched length, so account- or model-specific claims need fresh verification.

Why Older “One Minute” Answers Are Misleading

Older Sora announcements described research or earlier product capabilities. They should not be treated as the current API contract. For an implementation decision, the live API reference is the stronger source: it currently enumerates 4, 8, and 12 seconds for new clips and separately documents extensions.

Current-status note

Sora 2 is listed as deprecated

OpenAI's current model catalog labels Sora 2 and Sora 2 Pro as deprecated. Check the catalog and API reference again before planning a production workflow around either model.

View the official OpenAI model catalog →

Three Ways to Make a Longer AI Video

01

Extend a completed clip

Use a supported extension workflow when the next moment must continue from an existing video. Recheck current model access before committing to production.

02

Plan multiple focused shots

Break the story into setup, action, detail, and payoff. Generate each beat separately so one weak segment can be replaced without restarting the whole sequence.

03

Assemble the final sequence

Keep shot order, visible character details, format, and creative direction consistent, then finish timing, sound, and titles in your editor.

Sora Length vs AI Long Video

Sora is a generation model and API family. AI Long Video is a finished-video production service: 30 seconds uses one native Seedance 2.5 generation, while 60 seconds uses two connected 30-second scenes with native audio and automatic backend assembly.

See the exact 60-second workflow →

Compare AI Long Video vs Sora →

Frequently Asked Questions

How long can a Sora AI video be?

OpenAI's current video API reference lists 4-, 8-, and 12-second values for a newly generated clip. The API also exposes video extensions, whose returned duration is the stitched total, but the reference does not publish one universal maximum stitched length.

Can Sora extend an existing video?

The current OpenAI video API includes an endpoint for extending a completed video. Availability still depends on model and account access, and the Sora 2 model family is currently marked deprecated in the official model catalog.

Can Sora generate a one-minute video in one request?

The current API reference does not list 60 seconds as a base creation value. It lists 4, 8, and 12 seconds for new clips, so a longer result requires a different workflow such as extensions or assembling multiple shots.

Build beyond one short clip

Create a Finished Long Video

Choose 30 or 60 seconds and receive one playable, downloadable video without manual clip assembly.

Open the 60-second Studio →