Seedance 2.0 AI video generator

What is Seedance 2.0?

Seedance 2.0 is the second major generation of the Seedance video model family from ByteDance Seed, officially launched on 12 February 2026. Compared with the 1.x line it moved from short text or single-image clips to a multimodal system: you can drive a generation with a prompt, a first and last frame, several reference images, reference video clips and reference audio at the same time, and the model produces its own synchronized sound. People search for it as "seedance 2" or "seedance2.0"; it is one model, and this is its page.

Seedancee is an independent platform and is not affiliated with ByteDance. We provide Seedance 2.0 through Kie.ai's hosted endpoint (bytedance/seedance-2), keep a copy of every finished video in our own storage because provider links expire after 14 days, and read every capability shown on this page from a registry that we re-verify against Kie's documentation. The verification date is shown in the status strip above.

Create with Seedance 2.0

The generator below is preset to Seedance 2.0. Configure the prompt, mode, duration, aspect ratio, resolution and audio; the credit estimate updates as you change settings. Running the generation needs a subscription: Seedance 2.0 is part of the Pro plan, which renders it at up to 4K and the full 15 seconds. The Starter plan does not include it; on Starter, Seedance 2.0 Fast and 2.0 Mini are the closest options.

Seedance 2.0 capabilities on Seedancee

As exposed by Kie.ai on 29 September 2026, Seedance 2.0 outputs 480p, 720p, 1080p or 4K, runs 4 to 15 seconds (or automatic duration), and accepts up to 9 reference images, 3 reference videos and 3 reference audio files. Output is MP4 with optional native audio. The table places it next to Seedance 2.5 and Seedance 2.0 Mini so the trade-offs are visible.

Workflows

Text to video

A prompt alone is enough. Seedance 2.0 responds well to concrete camera language ("slow push-in", "low-angle tracking shot"), lighting notes and a sentence describing the sound. Because the ceiling is 15 seconds, one idea per generation works better than a shot list.

Image to video

Upload a first frame and optionally a last frame; the model animates between them and can return the final frame so you can chain the next shot. This is the workflow for product shots, real-estate stills and any case where the source must be preserved exactly.

Reference to video

Attach up to 9 images for identity and style, up to 3 clips for motion or camera language and up to 3 audio files for rhythm. Reference videos add a flat surcharge to the credit cost, shown in the estimator. For projects that need more references, Seedance 2.5 accepts up to 30 images, 10 clips and 10 audio files.

Edit and extend

Kie does not expose edit or extend for the seedance-2 endpoint, so those modes are unavailable for 2.0 on Seedancee. If you need to continue or revise a shot, generate it with Seedance 2.5 instead, or use the last frame of a 2.0 result as the first frame of a new generation.

Native audio

Seedance 2.0 generates audio together with the video: room tone, foley matched to on-screen actions and simple musical cues, from a description in the prompt. Put the sound in its own final sentence. Speech can be described but should be treated as texture rather than scripted dialogue; describe it as inaudible or in the background if lip-accurate lines are not the goal. Audio does not change the credit price on Seedance 2.0.

Example prompts for Seedance 2.0

A steel dive watch with a navy dial rests on wet black basalt. Extreme close-up on the crown, then a slow tilt across the dial as a beam of hard sunlight sweeps over the crystal and throws a caustic reflection onto the stone. Shallow depth of field, cool tones, subtle film grain. Sound: a distant wave, a single soft tick.

Starting from the uploaded photo of the living room, a slow steady push-in toward the floor-to-ceiling window as the afternoon light moves across the oak floor; no people, no camera shake, furniture stays exactly as pictured. Sound: quiet room tone, faint birdsong outside.

Using the referenced character images, the courier with the red scarf runs up a wet iron staircase at night, pauses at the top to catch her breath, then looks down at the city and grins. Handheld, low key, sodium-orange street light. Sound: footsteps on metal, rain, distant sirens.

Seedance 2.0 vs Seedance 2.5

Seedance 2.5, launched 31 July 2026, did not replace 2.0 on Kie; both are live and each has a job. Choose 2.0 when you need a 4K master for a 4–15 second shot. Choose 2.5 when you need more than 15 seconds (up to 30), many references, or the ability to edit and extend a result. Costs differ too: 2.0 has a lower credits-per-second base than 2.5 but a higher multiplier at 4K, so compare the estimator for your exact settings.

Release timeline and current status

• 12 February 2026: ByteDance launches Seedance 2.0.
• 31 July 2026: ByteDance launches Seedance 2.5; 2.0 remains available.
• 29 September 2026: Seedancee verifies the Kie.ai seedance-2 endpoint (480p to 4K, 4–15s, native audio, MP4) and the seedance-2-fast and seedance-2-mini variants.

Seedance 2.0 is available on Seedancee for text-to-video, image-to-video and reference-to-video. Two related endpoints share the same generation: Seedance 2.0 Fast (480p/720p, quicker turnaround) and Seedance 2.0 Mini (480p/720p, lowest cost). Both appear in the model picker and are described on the Mini page.

Free access, pricing and API

Seedance 2.0 is not free to run anywhere we can verify: Kie charges per generation, and ByteDance's own channels offer limited trials rather than unlimited use. Seedancee charges credits per generation, generating needs a subscription to the Pro plan, and there are no free trial credits. Pricing lists the plans and the estimator computes the exact reservation for any 2.0 configuration, including 4K. Seedancee's REST API (/api/v1) runs Seedance 2.0 with the same credits and automatic refunds as the workspace: it is part of the Pro plan and answers once API access is switched on, and until then it is in early access. The box below shows the current status.