ByteDance: Seedance 1.5 Pro

Seedance 1.5 Pro is a video generation model from ByteDance, listed on api14 as bytedance/seedance-1-5-pro. It belongs to ByteDance's Seedance line of audio-visual generation models, and the provider describes it as built on a 4.5B parameter Dual-Branch Diffusion Transformer architecture. It sits in the video category, so its role is generating video rather than handling text or chat tasks.

The model accepts text and image inputs and produces video as its only output. Because it generates video and audio simultaneously in a single unified pass, audio is part of the generated result rather than something added afterward. Output can be produced at 480p, 720p or 1080p, in aspect ratios of 1:1, 3:4, 9:16, 9:21, 4:3, 16:9 and 21:9, and at durations from 4 to 12 seconds in one second steps. It also supports frame images, specifically a first_frame and a last_frame, which lets you anchor the start, the end, or both ends of a clip. No context length in tokens is published for this model, which is expected for a video generator, and maxOutput is 0 in the model metadata.

Those capabilities point to a fairly specific set of uses. Text to video suits short conceptual clips, social formats and storyboard previews, especially since the aspect ratio list covers square, vertical and widescreen delivery. Image to video, combined with first_frame and last_frame, fits workflows where you already have a still or a keyframe and want motion between defined visual bookends, such as animating a product shot or bridging two frames of an edit. Native audio generation means you can request clips where sound is part of the output instead of scoring them separately. The 4 to 12 second range makes the model best suited to short shots rather than long continuous sequences, and the resolution options let you trade preview quality against final output quality.

You call it through api14's OpenAI-compatible API using the model id "bytedance/seedance-1-5-pro". Because the interface follows the OpenAI format, you can use the same client libraries and request structure you already use for other models on the gateway, pointing the model field at this id. Among request parameters, frequency_penalty is the one listed as supported for this model. Generation settings such as resolution, aspect ratio, duration, first_frame, last_frame and audio generation are exposed through the model's own options rather than the standard sampling parameters you would use for a text model.

It is worth checking your requirements against what the model exposes before committing. There is no published token context length, and maxOutput is 0, so this is not a text completion model and should not be used where you need token based prompting or textual responses. Its output is capped at 12 seconds per generation, so longer pieces need to be assembled from multiple clips. If your task is text, image understanding or another non video modality, a model in that category on api14 will fit better.