Google is improving Veo 3.1’s “Ingredients to Video” capability, which lets users generate videos based on a reference image.
Google’s Veo 3.1 also lets users upscale videos to 1080p and 4K resolutions
Photo Credit: Google
Google upgraded its Veo 3.1 artificial intelligence (AI) model with new capabilities and improved instruction following on Tuesday. In particular, the Mountain View-based tech giant is improving the “Ingredients to Video” feature, which lets users add one or more reference images to generate videos based on them. The company says that after the update, users will notice higher character and background consistency, while making it easier to get the desired output without having to type long, detailed prompts. However, the biggest addition is the native support for vertical videos that can be uploaded as YouTube Shorts or Instagram Reels.
In a blog post, the tech giant announced and detailed the changes made to its latest video generation model. These improvements now bring the AI model closer to Sora 2, which grew popular after OpenAI released it within the new Sora app. OpenAI's main focus was on character consistency and prompt adherence while using reference images or videos, and that is exactly where Google's update focuses.
Veo 3.1 comes with a feature dubbed “Ingredients to Video.” It is essentially the capability that allows users to generate an image-to-video with an additional text prompt for guidance. Users can upload one or more images and instruct the AI model to use them in the narrative of the generated video. Google says that several aspects of this feature have now been improved.
Google says Veo 3.1 can better understand prompts, eliminating the need to write long and detailed instructions. Now, even with short prompts, the AI model can generate videos with “richer dialogue and storytelling.” In an example, the prompt “Documentary style, a raccoon manages a coffee shop, dialogue” and an image of the said raccoon generated a cinematic video of the animal sitting behind the counter of a coffee shop and interacting with customers.
The update also improves the character consistency, meaning no matter the setting or the prompt, the character's appearance and facial details will not change and remain the same throughout the video. This feature is also helpful when a user wishes to stitch together multiple videos with the same character to create a longer story.
Additionally, the AI model offers a similar level of consistency to the setting and the objects placed in a scene. Google says this will allow users to reuse an object, background, or texture across scenes. Veo 3.1 also allows combining disparate elements into a cohesive clip, offering more creative freedom.
More importantly, with native support for the 9:16 aspect ratio, users can directly upload a generated clip to YouTube Shorts, Instagram Reels, or TikTok without having to worry about cropping or loss of quality. Further, the company is also improving the 1080p resolution quality and adding a new 4K resolution upscaling to let users export the output for larger screens.
Veo 3.1's Ingredient to Videos feature is now rolling out to YouTube Shorts and the YouTube Create app. It is also available in the Gemini app for end users. For enterprises and developers, the AI model is available in the Flow app, the Gemini application programming interface (API), Vertex AI, and Google Vids. However, the latter does not support 1080 or 4K video resolution.
Catch the latest from the Consumer Electronics Show on Gadgets 360, at our CES 2026 hub.