What's new
A running log of what actually shipped on Roma AI, newest first.
"Render Console" visual identity, site-wide
A full visual redesign across the homepage, every marketing/auth page, and every tool inside the app (Image, Video, Workflow, Dashboard, History, Credits, Affiliate, Admin) — dark "ink" background, an amber/cyan functional color pair (amber = in progress, cyan = done), monospace type for headings and data, and camera-viewfinder frame chrome replacing plain rounded cards. The new "R" glyph logo is unified everywhere.
Real per-generation pricing + refund audit tools
Video pricing now reflects duration, resolution/quality, and add-ons like audio as they’re changed, live, instead of one flat number — including full controls for the Seedance and Wan model families. Added an internal Refund Audit dashboard for catching failed or stuck video jobs that never got auto-refunded.
GPT-powered prompt enhancement
A Classic/GPT toggle on Image, Video, and Workflow model nodes: GPT can turn a rough idea into a well-structured prompt (and negative prompt) before it’s sent to the model, with Low/Medium/High quality tiers priced separately. I2V models even get to enhance based on the uploaded first-frame image, not just the text.
Affiliate Portal redesign
Reworked the in-app Affiliate dashboard: clearer empty states, tighter density, a real payout-progress bar toward the $20 minimum, and instant copy-link feedback.
Unified Dashboard
Replaced separate scattered views with one Dashboard: a stats bar (credits, 30-day spend, total generations, most-used model) plus a single filterable activity table combining image generations, video generations, and Workflow projects by recency.
Workflow Projects (multiple saved canvases)
Workflow moved from a single throwaway canvas to a proper "My Projects" system — save, name, reopen, and delete up to 5 named Workflow canvases, each autosaving as you build.
Admin controls: Suspend, Activate, Delete permanently
Account-management actions for the admin dashboard, plus a cleanup of how negative prompts are resolved on the Video page (one field, one model-aware default, no more double-application).
Roma AI launches
Public beta: AI image and video generation across 30+ models (FLUX, Nano Banana, HiDream, Recraft for images; Kling, Seedance, Luma Ray 2, Wan, CogVideoX for video), a credit-based pricing system, and the first version of the Workflow canvas.