Skip to main content
Advanced AI Studio offers a searchable model picker. Choose the model first: each model has its own accepted resolutions, aspect ratios, references and advanced controls. The picker and a successful quote are the final authority for a generation you can run. The standard Core generator uses a different, built-in service and does not offer the Advanced Studio model picker.

​
Image models

Image choices can include Create new / Edit, output ratio, resolution, image count, quality, file format and transparent background—only when the model supports that option. Some useful distinctions in the current model registry:
  • Gemini 3.1 Flash Image supports listed resolutions through 4K, with up to 10 reference images in the supported request.
  • GPT Image 2 and its Flare/Sunburst variants include supported quality, format and transparent-background choices; an opaque output or a PNG/WebP-compatible transparent output must match the selected settings.
  • FLUX 3 Image has a wider registered aspect-ratio range, resolutions through 4K, and a current maximum of two outputs per request.
  • Seedream 5.0 Lite/Pro support up to two images per request in the current registry, not the default image-count maximum used by some other models.
  • Muse Image is registered at 2K, with a conservative maximum of four reference images.
Reference-image capacity varies by model: 10, 8, 5 or 4 are examples from the registry. In edit mode, the image being edited counts toward the model’s reference budget.

​
Video models

These are registered request options, not a guarantee that every model/configuration is currently accepted by its provider.

​
First and last frames

Choose first frame to guide the opening image of a video, or last frame to guide its ending when offered. Some models support both; Grok Imagine Video currently registers only a first-frame slot, while Grok Imagine Video 1.5 registers both. The model picker should hide unsupported slots. Some models also accept other reference types such as images, videos or audio. The input format and count are specific to the selected model.

​
Audio, seed, aspect ratio and resolution

  • Generate audio appears for video models with registered audio generation. A model-generated soundtrack is different from the Video Combiner’s optional music track.
  • Seed / repeat number appears only where supported and can guide repeatability; it does not guarantee identical outputs.
  • Duration, aspect ratio and resolution are selectable only from model-supported options. A model may support 720p but not 4K, or portrait but not every widescreen ratio.
Image prompts have a 4,000-character registry limit; video prompts have a 10,000-character limit.

​
Choose the right workflow

  1. Select Image or Video and search by model or provider.
  2. Choose the target model before attaching frames or setting advanced options.
  3. Add references, media and model-supported settings.
  4. Review the current Creative Credit quote and approve the required maximum.
  5. Generate, then inspect the saved file and its delivery status.
Switching models may reset or invalidate options that the new model cannot accept. Request a fresh estimate after changing the model or references.

​
Model availability and pricing

Model registration is not a promise of universal account access or successful generation. Availability can depend on rollout, gateway support, provider acceptance and your exact settings. New and retired models can change independently of subscription limits. Credit cost is dynamic. The current wallet policy charges 2 customer Creative Credits per underlying provider credit; the interface can temporarily reserve extra headroom before the final settlement. Do not use a fixed per-model price taken from an old example. See AI content generation, Workflow Canvas and Creative Credits.
Last modified on October 20, 2026