Pencil is excited to be able to offer the Google Veo2 Video Generation model.
Configuration
In order to gain access to Veo2, a workspace admin will need to enable it by going to Settings > AI Governance and selecting it within the Video Generation drop down:
β
Google Veo2 is a Text to Video model that also supports Start frames, but you can still use Stable Diffusion if you require the End frame functionality.
Within the Video Controls you can configure your Aspect Ratio and Resolution (which can be enhanced later), a duration of between five and eight seconds and you are limited to generating one variation at a time. Finally is a text box for an optional seed value
Prompting in Veo2
The articles linked below contain more detailed information regarding prompting for video but suffice to say some key considerations apply:
The more detail, the better. Giving Veo2 as much rich detail as possible will ensure the highest chance of it generating content that aligns with your vision.
Specify and describe, in as much detail, as possible a subject, and their visual context and any background elements.
Describe in detail the movement you want within the frame - typically the subject but potentially background elements too. Google Veo2 is strong at interpreting movement/positional language.
Describe in detail any camera movement that you want (see here for tips)
Describe in detail the lighting that you want in your video - considering how your subject is lit and how that may contrast with the background.
Describe any specific style you want, if not realistic, which is the default (e.g. cartoon, animated, futuristic)
Include any visual themes and any other specific descriptors.
Include a negative prompt if desired.
Google Veo2 supports Seeds and it requires you to enter a seed value, which can be anything. It does not generate seed values for generated videos.