ByteDance launches Seedance 2.5 AI video generation model

Seedance 2.5 ByteDance AI video generation multimodal reference MiniMax H3

ByteDance announced the release of Seedance 2.5, an AI‑powered video generation model capable of creating single clips up to 30 seconds in length. The system builds on the multimodal audio‑video architecture of Seedance 2.0 and adds longer narrative capacity, multi‑round extension, and enhanced editing tools. Users can provide up to 30 images, 10 video clips and 10 audio clips as reference material, allowing the model to maintain consistent characters, scenes and sound across generated segments.

The rollout will appear first in Dream AI and Doubao Professional, with API access planned through Volcano Ark. Seedance 2.5 targets film, advertising, education, industrial manufacturing, embodied‑AI and autonomous‑driving workflows, aiming to move video creation from isolated clips to complete creative pieces.

In parallel, MiniMax released its H3 open‑source video model, which offers “Seedance‑level” multimodal control and AI‑native post‑production. H3 accepts mixed inputs (up to nine images, three video clips and three audio clips) and can produce 10‑second, 16:9 videos with precise textual instructions for scene sequencing, supporting use cases such as product demos, UI walkthroughs and short ads while lowering production costs.

Entities: ByteDance Ltd. · Dream AI · DreamView · MiniMax H3 · Schmidt Group · Seedance 2.5 · Seedream 5.0 Pro · Volcano Ark

Claims

What the coverage asserts, and how well corroborated each claim is across sources.

[● 2 SOURCES] Seedance 2.5 supports multi‑round extension to continue videos while preserving scene and character consistency. (Seedance 2.5 features)
[● 2 SOURCES] ByteDance released Seedance 2.5 video generation model on July 31. (ByteDance announcement)
[● 2 SOURCES] Seedance 2.5 accepts up to 30 images, 10 video clips and 10 audio clips as reference input. (Seedance 2.5 multimodal capabilities)
[○ 1 SOURCE] MiniMax H3 can generate 10‑second 16:9 videos with precise textual scene sequencing. (MiniMax H3 capabilities)
[○ 1 SOURCE] MiniMax H3 can ingest up to nine images, three video clips and three audio clips (total 12 inputs). (MiniMax H3 input limits)
[● 2 SOURCES] Seedance 2.5 can generate single video clips up to 30 seconds long. (Seedance 2.5 specifications)
[○ 1 SOURCE] Seedance 2.5 will initially appear in Dream AI and Doubao Professional, with API access via Volcano Ark. (ByteDance rollout plan)
[○ 1 SOURCE] MiniMax H3 offers Seedance‑level multimodal control and AI‑native post‑production capabilities. (MiniMax H3 product description)