Wan 2.5 API: Video and Audio in One Pass

The Wan 2.5 API gives developers Alibaba's Wan 2.5 video model on Atlas Cloud through one OpenAI-compatible key. Its defining trait is one-pass audio-visual generation: it produces the picture and a synchronized soundtrack, including voice, sound effects, and lip-sync, in a single step. Built on a Diffusion Transformer architecture, it runs alongside 300+ other models on the same account.

Yes. The Wan 2.5 API creates video and its audio together in one pass, aligning voice, ambient sound, and lip movement without a separate recording or manual sync step. A single structured prompt returns a finished clip with sound already in place, which removes the record-and-align stage that most video pipelines still need.

Wan 2.5 generates at 480p, 720p, and full 1080p, with clip lengths of 5 or 10 seconds. The tiered resolutions let you draft cheaply at 480p and deliver at 1080p from the same request. Exact resolution and duration options depend on the mode and configuration on Atlas Cloud, so confirm the current settings in the console before building around a specific output.

The Wan 2.5 API covers text to video and image to video, with a speed-optimized Fast variant for image to video and a text to image mode. Text to video builds a scene from a prompt, while image to video animates a still and preserves its identity and style. Switching modes is a change to the model name on one integration.

Yes. Alongside generated sound, the Wan 2.5 API accepts an uploaded audio file to drive a clip, so the video can follow a specific voiceover or track. It also reads cinematic direction from the prompt, such as pans, tilts, zooms, and dolly moves, giving control over both the soundtrack and the camera work in a scene.

Wan 2.5 keeps voice and lip movement aligned across languages and reliably processes Chinese prompts for audio-visual generation, an area where some competing models fall back to an unknown-language response. This makes it dependable for localized dialogue and cross-market content where the audio has to match the spoken language on screen.

The three generations serve different needs. Wan 2.2 is the open, Mixture-of-Experts model for those who want open weights; Wan 2.5 centers on one-pass audio-visual generation with 480p to 1080p tiers; and Wan 2.6 focuses on multi-shot storyboard storytelling with character consistency. Because all three sit behind one key on Atlas Cloud, you can pick the version that fits each job.

Fast is a speed-optimized Wan 2.5 variant for image to video, tuned for lower latency. Use it for drafts, batch generation, and prompt testing where quick turnaround matters more than maximum polish, then move approved shots to the standard model for the final render. It keeps iteration quick without changing your integration.

Yes. Video generated with Wan 2.5 can be used in commercial work such as ads, product demos, and published content. Review Atlas Cloud's terms of service for the specifics of your plan, and note the usual restrictions around generating content that depicts real, identifiable people without their consent.

Create an account on Atlas Cloud, generate an API key, and send a request to the Wan 2.5 model with your prompt or input image through the OpenAI-compatible endpoint, choosing your resolution and duration. Poll the prediction endpoint for the finished clip, then scale up as needed. The same key reaches 300+ other models, so you can test alternatives without extra setup.