Can Sora make videos from images?
Yes, Sora supported image-to-video: you uploaded a starting photo and it animated it into a short clip. It also allowed uploading images of people (with consent) for cameos. This is now historical, though — OpenAI discontinued the Sora app and website on April 26, 2026.
Why — the first-principles explanation
Turning a still image into video works because of how Sora generated frames. Instead of starting purely from random noise, the model could use your image as the first frame — an anchor — and then predict what the following frames should look like, extending the scene into motion. It leaned on everything it learned about how objects and light behave over time to imagine a plausible continuation.
Sora offered two flavors of this. Basic image-to-video took a starting frame — say, a landscape photo — and brought it to life with camera movement and motion. The Cameo feature was a special case: you (or someone who gave consent) uploaded footage or images of a real person, and Sora could then place that likeness into new scenes. Because a person's face is involved, that path required explicit consent and could be revoked at any time.
The realism of this depended on the prompt you paired with the image. A clear instruction — 'slowly zoom out as clouds drift past' — guided the motion better than the image alone. As with everything Sora made, outputs carried a visible watermark and C2PA provenance metadata. And the practical caveat for 2026: the consumer app is gone as of April 26, 2026, so image-to-video now lives in other tools like Runway, Kling, and Google Veo.
An example that makes it click
Imagine handing an artist a single photo of a calm lake and saying, 'now show me the next ten seconds.' The artist studies the picture, then paints frame after frame of ripples spreading and a bird flying across — inventing the motion that logically follows from your one still image. That's what Sora did: your photo was frame one, and it dreamed up the rest.
How to do it
- Note the status: Sora's app closed April 26, 2026, so these steps describe how it worked.
- When live: open Sora and choose the option to upload an image as your starting frame.
- Add a text prompt describing the motion you want, such as camera movement or action.
- For a person's likeness (cameo), confirm you had consent before uploading.
- Generate, review, and download the clip (it carried a visible watermark).
Key facts
- Sora supported image-to-video, using an uploaded photo as the starting frame.
- The Cameo feature let users animate a real person's likeness, with consent.
- Pairing the image with a clear text prompt improved the resulting motion.
- Every output carried a visible watermark and C2PA provenance metadata.
- The consumer app was discontinued April 26, 2026.
OpenAI's text-to-video model for short cinematic clips.
Affiliate link — we may earn a commission at no cost to you.▶ The 60-second explainer (script)
Can Sora make videos from images? Yes, it could. This was called image-to-video. You uploaded a photo, say a still of a quiet lake, and Sora used it as the very first frame, then imagined the motion that would naturally follow, like ripples spreading and clouds drifting by. It figured this out from everything it learned about how the world moves. There was also a special version called Cameo, where you could upload footage of a real person, with their consent, and place them into new scenes. To get good results, you paired your image with a clear prompt describing the motion. Every clip came out with a watermark showing it was AI-made. One catch for today: the Sora app shut down on April 26, 2026, so for image-to-video now, you'd use tools like Runway, Kling, or Google Veo.
What authoritative sources say
People also ask
Did Sora animate any photo?
It could animate most uploaded images, but photos of real people required consent through the cameo system.
What is image-to-video?
A feature that uses your uploaded picture as the first frame and generates motion that follows from it.
Did image-to-video clips have watermarks?
Yes. Like all Sora output, they carried a visible watermark and hidden C2PA provenance metadata.
What tools do image-to-video now?
With Sora gone, active options in 2026 include Runway, Kling, and Google Veo.