Video generation is one of the fastest-growing fields in AI. In just the last couple of years, text-to-video tools have evolved from curious clips with blurry faces to quite convincing short scenes. And this progress continues: Wan 2.7, a suite of four models each responsible for a specific type of video task, recently became available on the Together AI platform.
Four Models, Four Tasks
Wan 2.7 isn't a single universal model, but a complete suite. Each of the four models was created for a specific task: generating video from a text description, continuing an existing video, working with reference images, and editing finished material.
Simply put, while earlier video generation tools could mainly do one thing – create something from scratch based on text – Wan 2.7 covers the entire basic workflow: create, continue, guide with a sample, and edit.
The launch began with the «text-to-video» model, which was the first to become available on the platform. The other three will follow as the rollout proceeds.
Why Use a Reference – and What Is It Anyway?
One of the most interesting modes in the suite is working with references. In short: instead of describing everything with words, you can show the model an example – an image or a video clip – and ask it to create new content in the same style, with the same characters, or in a similar setting.
This is especially important for tasks that require visual consistency. For example, if you are creating a series of videos with a single character, the reference mode allows you to maintain their appearance from scene to scene without having to write detailed text descriptions each time.
Editing as Part of the Workflow
The editing model deserves special attention. This isn't about editing in the traditional sense, but about changing the content of an existing video using AI – replacing objects, backgrounds, styles, or individual elements of a scene.
This area is currently being actively developed by many teams because the demand is clear: editing existing material is often more convenient and cheaper than generating everything from scratch. Including such a tool in a unified suite is a logical step.
Why It's Appearing on Together AI
Together AI is a developer-focused platform: it provides access to open and partner models via an API, without the need to deploy one's own infrastructure. Wan 2.7 is appearing here as part of this logic – to give teams quick access to video tools without forcing them to set up an environment from scratch.
For developers already working with the platform, this means they can integrate video generation into their products through a familiar interface – without switching tools or creating separate integrations for each model.
What This Means in a Broader Context
Wan 2.7 isn't a revolution, but it is a telling sign of the field's maturity. Video generation is gradually moving away from being a collection of disparate experimental tools and is starting to form into coherent workflows.
Four models under one brand, covering the full cycle from creation to editing, reflects a product-oriented mindset rather than just a demonstration of technical capabilities. This approach is convenient for both solo developers and teams that need to integrate video into more complex products.
For now, the rollout is just beginning with the text-to-video model. How practically useful the other three will be will become clear as they appear on the platform.
