Seedance 2.5 is ByteDance Seed's audio-video generation model for longer narrative sequences, multimodal referencing, and targeted editing. ByteDance documents a 30-second single-pass maximum, multi-round video extension, timestamp-level editing, and synchronized audio-video generation. The official reference workflow can accept up to 30 images, 10 video clips, and 10 audio clips in one request.
Those provider-level capabilities are broader than the integration currently shown on this GPTProto model page. The live GPTProto Seedance 2.5 API page exposes text-to-video generation from 4 to 15 seconds, with controls for aspect ratio, resolution, generated audio, fixed-camera behavior, and seed. Developers should build against the values in the live endpoint rather than assuming every capability announced by ByteDance is already available through this model ID.
Seedance 2.5 Specifications on GPTProto
| Specification | Verified current value |
|---|---|
| Provider | ByteDance Seed |
| GPTProto model string | dreamina-seedance-2-5-260628 |
| Current GPTProto task | Text-to-video |
| Current GPTProto duration | 4–30 seconds |
| Current page default | 5 seconds, 16:9, 720p, generated audio enabled |
| Exposed controls | Prompt, aspect ratio, duration, resolution, generated audio, camera fixed, and seed |
| Seed behavior | -1 is the current default; set an explicit value for repeatable tests |
| Official model maximum | Up to 30 seconds in one pass; not exposed by the current GPTProto duration control |
| Official reference capacity | Up to 30 images, 10 videos, and 10 audio clips; not exposed on the current GPTProto text-to-video page |
| Output | Video with optional generated audio |
Seedance 2.5 API Access on GPTProto
GPTProto is most useful when the application needs more than one model. One API key and one balance can cover Seedance 2.5, Seedance 2.0, Seedance 2.0 Mini, Kling 3.0 4K, and other models in the video model catalog. This reduces separate provider accounts and credit balances when a team is testing quality, cost, or fallback behavior.
Use dreamina-seedance-2-5-260628 for the text-to-video task documented on this page. Do not assume it is a drop-in replacement for BytePlus ModelArk: the model string, request schema, supported duration, and available modes differ. The live GPTProto parameter panel and fixed Quick Start block are the source of truth for requests sent through GPTProto.








