Upscale

• The upscaler has been migrated to real esrgan -- further enhancements are planned there over the coming weeks.

Inpainting

• Inpainting has been restricted to Venice PRO users while we diagnose stability and performance issues with the feature.

API

• The /api/v1/models endpoint will now return only text models by default. The type query parameter has been extended to support all | image | text and the type parameter has been added to the body to indicate what filter was used.

• The model list endpoint now includes a modelSource parameter:

{
            "id": "llama-3.3-70b",
            "model_spec": {
                "modelSource": "https://huggingface.co/meta-llama/Llama-3.3-70B-Instruct",
                ...
            },
            ...
        }

API Page

• API page now lists which tier your account is operating under (explorer vs. paid).

Misc

• Fixed a bug preventing safe venice in the image settings from properly toggling.

• Added a guard to prevent gifs from being uploaded to in-painting.

• https://chat.venice.ai will now redirect to https://venice.ai

Encrypted Backups Beta

• Encrypted backups has been released in Beta to our Venice Beta Tester group. If you'd like to help Beta Test, reach out to one of our moderators in #product channel on discord.