June 15, 2026
Media library and references
Save strong images and videos to a shared library, then bring them back as references so characters, products, and style stay consistent.
Useful AI results are often lost because they live inside a generation history or a downloaded folder. When the next project starts, creators waste time trying to recreate a frame they already had.
The Givon AI library is shared across projects. Save a strong image, video, product frame, character frame, voice, or motion reference once, then bring it back into any new scene.
References work best when each one has a clear role. One reference can define a face, another can define product shape, another can define lighting or motion. Clear roles make generation more stable.
This matters most for repeatable work: product content, AI characters, brand campaigns, client revisions, and ongoing social series.
What it gives you
The shared library is available across projects.
Saved images and videos can return to new scenes as references.
Uploaded source files can be stored alongside generated assets.
Reference roles help preserve characters, products, style, and movement.
Clean final takes can become start frames or source material for future scenes.
When it is useful
Series in one style
Reuse the same characters, products, frames, and lighting across episodes.
Brand and product content
Keep approved products, logos, backgrounds, and compositions close to the editor.
AI characters
Save stable portraits and reference frames so identity does not drift.
Faster new scenes
Start from working references instead of rebuilding every prompt from scratch.
FAQ
What should be saved to the library?
Save assets with a clear future role: character portraits, product frames, motion references, clean final takes, backgrounds, or voice assets.
Is the library project-specific?
Project materials stay inside each project. The library is shared across the account for assets that should be reused.
Can too many references hurt a generation?
Yes. References work better when each one has a clear role and the set does not give the model conflicting instructions.