Changelog
API Docs
Help Center

July 7th, 2025

App

Models

Token

Venice.ai Change Log - July 1st - July 7th, 2025

Image Editing (Inpainting) is Back!

We’ve brought image editing (inpainting) back to Venice for Pro users, now powered by Flux Kontext DEV.

How Image Editing it works
We’re running Flux Kontext DEV with a custom implementation that understands natural language editing commands. Here’s how it works:

  1. Upload an image using the paperclip button and select “Edit Image”

  2. Or generate an image first, then click the paint roller icon in the action menu

  3. Describe your edit in plain language,

Capabilities of the new Image Editing

  • Inpainting: Remove or replace objects (“remove the vase on the table”)

  • Style changes: Transform artistic styles (“change this to Ghibli style”, “make it monochrome”)

  • Compositional editing: Modify poses, angles, compositions (“create a variation of this image”, “rotate the person by 80°“)

  • Color/tone adjustments: Change lighting, colors, moods (“colorize this black and white photo”, “add dramatic shadows”)

We need your help testing the new Image Editing
Please try the feature and send any feedback, bugs, or unexpected behavior to [email protected].

A wider release along with API support is planned once we’ve addressed community feedback.

App Updates

  • Improved reliability of the simple mode feature in chat by ensuring the latest selected setting is always used, preventing issues caused by outdated values during asynchronous operations.

  • Improved performance of the Venice “Loading” keys.

Model Updates

  • Updated Venice Uncensored to an improved 1.1 version with some key improvements and bug fixes:

    • Better tone control: The model is now more neutral and polite by default, but still capable of being rude or explicit when prompted.

    • Improved ethical reasoning The model now handles ethical dilemmas better, prioritizing consequentialist reasoning to reduce bias and inconsistency.

    • Bug fixes: Issues like the model responding inappropriately to minimal inputs (e.g., single letters) have been addressed.

  • Removed Qwen Coder and DeepSeek Coder after deprecation period.

Token Updates

  • Updated all references from Venice Compute Units (VCU) to Diem - See this blog for more details.

  • Fixed duplicate toasts being displayed when a transaction fails.