Changelog

Free AI Model (text-to-image) Added a free text-to-image model based on flux-2-dev, exposed as “Free AI Model” in the model selector. Cost is 0 credits, fully integrated with the unified credit-cost system. Free AI Model (image-to-image, daily limited) Added z-image-turbo as a free image-to-image model with 0 credit cost. Introduced a per-user daily limit (10 calls/day) on the backend using getDailyModelUsage, returning HTTP 429 when exceeded. New model variants and pages Nano Banana 2: new model wiring in TEXT_TO_IMAGE_MODEL_MAP / IMAGE_TO_IMAGE_MODEL_MAP and UI model lists; dedicated /models/nano-banana-2 page plus a comparison section on the original Nano Banana page. Seedream v5: new Bytedance Seedream v5 text-to-image and edit endpoints, wired into both text-to-image and image-to-image model lists and a dedicated /models/seedream-5 page. Qwen Image Max: upgraded from the previous Qwen Image endpoint to qwen-image-max with corresponding model mapping and a new /models/qwen-image page focused on Qwen Image Max. Free AI Model landing page: dedicated /models/free-ai-model page with FAQ content.

Per-model credit pricing everywhere Frontend (text-to-image & image-to-image pages) now uses MODEL_CREDIT_COST[model] instead of a hard-coded “1 credit”. Buttons display dynamic labels such as “Free”, “1 Credit”, or “N Credits”, and disable state now checks credits < creditCost rather than credits <= 0. GPT-Image-1.5 keeps its quality-based pricing (low/medium/high → 0/1/3 credits) but now flows through a unified helper. Backend credit handling generate-image API now: Skips credit balance checks and deductions when creditCost === 0. Applies the daily free-limit only for z-image-turbo (image-to-image) with a hard limit of 10 calls per day per user. Keeps recordImageGeneration for all models (including free ones) but only deducts credits when creditCost > 0. Special-cased payload handling for z-image-turbo vs other image-edit endpoints (single image_url vs image_urls[]; conditional aspect_ratio / GPT-specific fields). ImageCombiner UX and model list logic ImageCombiner now: Uses TEXT_TO_IMAGE_MODEL_LIST for text mode and IMAGE_TO_IMAGE_MODEL_LIST for image-edit mode, with useMemo for derived state. Automatically fixes selectedModel when switching modes if the current model is not present in the new list. Centralizes the “Run” button credit label via runButtonCreditText, reusing MODEL_CREDIT_COST and GPT-Image-1.5 quality rules. Switched internal image preloading to document.createElement('img') for better consistency. Model catalog & navigation Updated footer model list: Added “Free AI Model”, “Nano Banana 2”, “Seedream v5”. Renamed Qwen entry to “Qwen Image Max”. Reactivated the homepage announcement bar, now pointing to /models/nano-banana-2 with “Nano Banana 2 is now live!”. Qwen model references across the app now consistently use “Qwen Image Max” and qwen-image-max. UI polish Improved model <Select> dropdown width/positioning so it respects trigger width and viewport on small screens, both on the text-to-image page and in ImageCombiner. Button credit badges on mobile/desktop are now consistent and dynamically reflect the current model and parameters.

Nano Banana Pro prompt generator released - Generate prompts for Nano Banana Pro

ChatGPT style image generator released - Transform your photos into beautiful ChatGPT anime-style artwork

iOS Emoji icon generator released - Create iOS emoji-style icons

Background remover tool released - Remove background from your photos

Studio Ghibli style image generator released - Transform your photos into beautiful Studio Ghibli anime-style artwork

Enhanced image-to-image generation with optimized model selection

login system released

payment system released

style transfer feature released

Endless creativity feature released

text-to-image feature released

image-to-image feature released