Purpose
Krea 2 is aimed at creative users who want images with an "authorial feel" rather than a generic style, and it is suited to visuals that need grain, texture, expressive lighting, and unconventional composition. Its official positioning is to keep artistic decisions in the user's hands, achieving results by choosing the model tier, writing clear art direction, and comparing controlled variables, rather than relying on luck in a single generation.
Main Capabilities
- Three Krea 2 tiers: Turbo for rapid exploration, Medium for everyday creation, and Large for scenarios with the highest demands for fidelity and raw detail, with all three sharing the same aesthetic direction.
- Multi-model route switching: In the same session, you can open the GPT Image 2 route for instruction-driven generation and editing, the Nano Banana route for Gemini-style rapid exploration, and the FLUX route for realistic control; when switching engines, prompts, reference materials, and generation history can carry over.
- Text and image input: You can start from a text brief, or upload reference images for style repainting and style-guided composition (the image-to-image route).
- Credits shown before running: An exact credit estimate is displayed before each generation, and failed generations are automatically refunded.
- History: Every result is saved to history along with its prompt and settings, making it easy to reproduce successful directions and derive variants.
- Asset format support: Images JPG, PNG, WebP, up to 20 MB; videos MP4, WebM, MOV, up to 100 MB; audio MP3, WAV, M4A, FLAC, up to 50 MB.
How to Use
Official generation workflow has four steps:
- Define the subject: First write clearly what the image is specifically meant to present, the thing or scene that must be preserved in all subsequent style decisions.
- Guide the image: Add composition, viewpoint, lens language, lighting, color tone, and texture; concrete physical descriptions are better than piling up style keywords.
- Generate and compare: Choose the model tier, check the credit cost, then run; change only one variable at a time so you can tell which wording improves the image.
- Save the recipe: Results enter history along with the prompt and settings, making it easy to reproduce and generate variants.
It also states: only routes available online can be selected, and the prompt remains unchanged when switching; after uploading the required assets, the generator will use them automatically; for the API, you can create API keys, submit generation tasks, and use the same account credit balance.
Use Cases
Examples of the official prompt structure include: character direction (fantasy subjectcombined with a flat, high-energy composition), an extreme low viewpoint with fisheye distortion turning a familiar tree into a monumental environmental image, and editorial-style action (close-range low-angle motion imagery, expressing movement, lens energy, reflective materials, and readable art direction). When the deliverable requires typographic precision or realistic control, you can switch to the GPT Image 2, Nano Banana, or FLUX route.
Limitations
- For specific pricing amounts, subscription tier names, and capacity details, the official information only provides descriptions such as "credit consumption is shown before generation," "monthly subscription or one-time top-up," and "some tiers include monthly credits, which can roll over, with up to 5x credit expansion"; specific prices and plan contents are not yet provided.
- Users must verify applicable laws, third-party rights, and relevant terms when creating paid content themselves.
- The official website states that "capacity uses this site's lowest-cost online configuration," while actual credit consumption varies with the selected model and settings.
Other
The official website also lists model catalog entries: GPT Image 2, Nano Banana, Nano Banana Pro, Nano Banana 2, FLUX.2 Max, FLUX.2 Pro, FLUX.1 Kontext, and Flux AI image generator, all marked as openable.
