Seedance 2.5 is now available in VlogMe through direct ModelArk

Seedance 2.5 is live in VlogMe

VlogMe now submits Seedance 2.5 directly to BytePlus ModelArk with the verified model identifier dreamina-seedance-2-5-260628. The current Studio launch covers text-to-video and one-image reference generation at 720p for 4–30 seconds with native audio.

The launch uses the same owned ModelArk key, task submission, callback, polling, billing, and storage path as the existing direct Seedance integration. New Seedance jobs do not pass through fal or Replicate.

What is available

Choose Seedance 2.5 in Text to Video for a prompt-only scene, or in Image to Video to supply one visual reference. The image is sent with ModelArk's verified reference_image role. Prompts can describe timed action, camera movement, dialogue, ambience, and a clear final composition across a clip of up to 30 seconds.

The current VlogMe route does not expose video references, audio-reference uploads, local editing, video extension, or first/last-frame control. Use Seedance 2.0 when its shorter workflow or first/last-frame option is a better fit.

Pricing and failure behavior

VlogMe calculates the Seedance 2.5 video portion from the verified 720p ModelArk token formula and the conservative no-video-input rate of $10.70 per million tokens. A five-second text-to-video request is 108,000 output tokens before VlogMe credit normalization. Image-to-video also includes the bounded trusted Seedream preparation used by the direct Seedance path.

Seedance 2.5 has no hidden cross-provider fallback. Authentication, schema, network, provider, and content-policy failures remain attached to the selected model and can be retried after the cause is resolved. The reviewed Kling safety fallback remains limited to the four compatible Seedance 2.0 public model IDs.