ByteDance: Seedance 2.5

ByteDance: Seedance 2.5 is a video generation model from ByteDance, listed on api14 under the model id bytedance/seedance-2.5. It sits in the video category, and ByteDance positions it for long-form storytelling, multimodal reference-based generation, video editing, and video extension. It is a generative media model rather than a text model, so the output you get back is video.

On the input side, Seedance 2.5 accepts text, images, video, and audio. On the output side, it produces video only. It supports first-frame control and first-and-last-frame control through the frameImages options first_frame and last_frame, and it can generate audio alongside the video. Output resolutions are 480p and 720p. Aspect ratios cover 16:9, 4:3, 1:1, 3:4, 9:16, and 21:9. Durations run from 4 to 30 seconds in single-second steps, so you can request anything from a 4 second clip up to a 30 second one. No context length in tokens is published for this model.

Those capabilities point to a fairly specific set of jobs. Multimodal input plus reference-based generation suits work where you want the output to follow a supplied image, clip, or audio track rather than a text prompt alone. First-frame and first-and-last-frame control is useful when a shot has to start or end on an exact image, for example in continuity work or when stitching generated clips into a longer sequence. The long end of the duration range, up to 30 seconds, fits narrative beats that would otherwise need several shorter generations joined together. Video editing and video extension make it a candidate for lengthening or reworking existing footage, and the audio generation option means a clip can come back with a soundtrack rather than silent. Multiple aspect ratios let the same model serve landscape, square, vertical, and wide cinematic framing.

To call it, send a request through api14's OpenAI-compatible API using the model id bytedance/seedance-2.5. Because the API is OpenAI-compatible, you can point an existing OpenAI client at api14 and change the model field rather than rewriting your integration. Among the parameters api14 lists for this model, frequency_penalty is supported; the rest of the request shape follows the standard endpoint you are already using for other models on the gateway. Check the live price table on the page rather than assuming a cost, since rates are not stated here.

One thing to weigh is that Seedance 2.5 returns video, not text, so it is not a substitute for a language model in a text pipeline; if your task is captioning, summarising, or reasoning over content, a text model on api14 is the better fit. Within video work, the choice depends on your target: 480p and 720p are the listed resolutions, so if you need something outside that range this model may not suit. The duration ladder starts at 4 seconds, so very short clips below that are not covered.