Seedance vs Sora 2: specs side by side after the shutdown

Seedance vs Sora 2 is now a comparison with a model you can no longer call: OpenAI shut down the Sora 2 models and its Videos API on September 24, 2026. On specs, Sora 2 made clips of up to 12 seconds in its API reference (20 in its guide) with synced audio, from text or one image, and could extend, edit and reuse characters; Seedance 2.0 makes 4–15 second clips and Seedance 2.5 up to 30 seconds, with optional sound, from text, first and last frames, and image, video or audio references.

Sora specs come from OpenAI's own Sora 2 and Sora 2 Pro model pages, its video generation guide and API reference. Seedance specs come from Sume's Video generation docs and its catalog. All were read 2026-09-28. What each cost is in Sora 2 API pricing.

How do Seedance and Sora 2 specs compare?

Sora 2 columns are what OpenAI's pages still document about lengths, frame sizes and inputs:

What did Sora 2 do that Seedance on Sume doesn't?

Three things in OpenAI's guide have no Seedance equivalent on Sume:

• Extensions: each added up to 20 seconds, up to six times, for 120 seconds in total. On Sume, one Seedance clip stops at 30 seconds; longer videos are several clips joined, as AI video length limits by model explains.
• Edits: Sora took an existing video and a prompt. The Seedance ids don't edit video; on Sume that is gemini-omni-flash-1.1.
• Characters: reusable non-human subjects uploaded as short clips. Sume's docs list no saved-character asset for Seedance; reference images or videos guide each clip instead.

What does Seedance take that Sora 2 didn't?

Sora's input_reference was one image, used as the first frame and matched to the output size. The Seedance ids take a first and a last frame, and references as images, videos or audio, which guide a fresh clip without being frames. Sora 2 generated video with synced audio; on Seedance, sound is optional through generate_audio.

Where to move a Sora integration, and how the old request fields map, is in Sora API alternatives.

How do Kling and Grok Imagine compare with Sora 2?

On specs alone: kling-3 makes 4–15 second clips from text or an image, with an end frame and optional sound, so it covers Sora 2's text and first-frame modes but not extensions or edits. grok-imagine-video-1.5 only animates one image into a silent 4–15 second clip, so it matches Sora's image mode without the audio. Kling vs Grok Imagine sets the two side by side.