Model Releases
Seedream 5.0 Pro and Gemini Omni Flash 1.1 Are Now in Karya

Two new models are live in Karya as of September 3, 2026: Seedream 5.0 Pro, an image model built for precise, reference-heavy editing, and Gemini Omni Flash 1.1, Google's video model with first and last frame control plus a video-input mode for extending or restyling an existing clip. Both are available in their respective studios, in Karya Flow, and from the mobile Create screen.
Seedream 5.0 Pro: precision editing with up to 10 references
Seedream 5.0 Pro handles both text-to-image generation and image-to-image editing, and it accepts up to 10 reference images in a single generation - enough to hold a product shot, a background plate and several style references at once instead of collapsing them all into one description. It is strong at photoreal detail, small on-image text, and precise region edits: point or lasso-style selections let you recolor an object, swap a material, or fuse elements from multiple references without disturbing the rest of the frame.
- Two quality tiers - Basic for fast drafts, High for the final pass.
- Six aspect ratios - 1:1, 16:9, 9:16, 4:3, 3:4 and 21:9.
- Point and lasso-style region edits that change only the selected area.
- Material and colour replacement without redrawing the whole image.
- Multi-image fusion from up to 10 references in one generation.
Gemini Omni Flash 1.1: video with first/last frame and video-input control
Gemini Omni Flash 1.1 covers three ways into a video: text-to-video from a prompt, image-to-video with both a first frame and a last frame so you can pin exactly how a shot starts and ends, and a video-input mode that takes an existing clip and extends or restyles it instead of starting from scratch. It renders at 360p, 720p, 1080p or 4K, in clips of 4, 6, 8 or 10 seconds, at 16:9 or 9:16.
Where to find them
Seedream 5.0 Pro is in the Image Studio for both generation and editing. Gemini Omni Flash 1.1 is in the Video Studio, covering text-to-video, image-to-video and video-input modes. Both also have nodes in Karya Flow, so you can chain a prompt or an LLM text node into either model, or fan a single setup out across a List node to batch variations. Both are reachable from the mobile Create screen too, so a phone is enough to start a generation.
Pricing
Karya shows the exact credit cost before you generate, and unused credits never expire.
- Seedream 5.0 Pro, Basic quality: 7 credits per image.
- Seedream 5.0 Pro, High quality: 14 credits per image.
- Seedream 5.0 Pro, edit: 9 credits per image.
- Gemini Omni Flash 1.1, up to 1080p: 55 credits (4s), 74 credits (6s), 92 credits (8s), 110 credits (10s).
- Gemini Omni Flash 1.1, 4K: 129 credits (4s), 147 credits (6s), 165 credits (8s), 184 credits (10s).
- Gemini Omni Flash 1.1, video-input mode: 147 credits (up to 1080p) or 221 credits (4K), regardless of length.
What to try first
If you already have a product shot or a brand reference set, start with Seedream 5.0 Pro: feed in your existing references and use High quality for anything client-facing, Basic for quick drafts. For video, image-to-video with a first and last frame is the fastest way to get a controlled result - pin the opening and closing look and let Gemini Omni Flash 1.1 fill in the motion between them. Save video-input mode for when you already have a clip you want to extend into a longer take or restyle into a different look.
Try it in Karya
Generate images and videos with frontier AI models from one canvas. No subscription - packs from $3, 200 bonus credits with your first purchase of $7+, and credits never expire.
Start free