GPT-6 Astra Still Can't Natively Generate Video

A capability breakdown of GPT-6 Astra shows the model remains a text-in, text-out system that can call an image tool, but has no built-in support for audio or video generation.

Confusion has been building around what GPT-6 Astra can actually do with images, audio, and video, and a new capability rundown from Elser AI tries to settle it by going straight to the model's official specifications. The short version: Astra itself is a text model. It reads text and images as input and answers in text. It does not natively produce audio or video, no matter how the surrounding app is marketed. The wrinkle that fuels most of the online disagreement is that Astra can be hooked up to an image-generation tool through OpenAI's Responses API. That means a product built on top of Astra can hand back a generated image, but that image isn't coming from Astra's own rendering ability, it's coming from a separate tool the model is simply calling. Elser AI frames this as three distinct and often conflated claims: a model understanding images, an app generating an image, and a model outputting video are not the same thing, even though headlines tend to blur them together. According to the capability breakdown, Astra supports text input and output along with image input and image-generation tool calls, but its official model page lists native audio input/output and native video input/output as unsupported. For anyone hoping to use Astra directly as a video generator, the advice is to look toward dedicated video and animation systems instead, since Astra was not built to handle that modality on its own. The report also notes an odd loose end: the same model page references endpoints for images, video, and audio elsewhere, even as the core modality table marks video and audio as unsupported. That inconsistency is likely part of why claims about Astra's abilities keep circulating in different, sometimes contradictory forms. For the AI video community, the takeaway is fairly grounded. GPT-6 Astra is not a new native video generator hiding in plain sight, it's a text-and-image reasoning model that can be wired into other tools, including image generators, through the Responses API. Anyone expecting it to output audio or video clips directly is going to be working against the model's documented limits, at least as they stand right now.