Interface

Deep Mirror – Basic & Advanced Modes

DeepMirror II presents two distinct interface modes, accessible via tabs at the top of the application: BÁSICO (Surface Mode) and AVANÇADO (Deep Mode). This dual-layer architecture addresses a fundamental challenge in creative tools: providing immediate, intuitive access for exploratory use while retaining full parametric control for expert manipulation.

Basic Mode — The Friendly Layer

Basic Mode functions as an accessible entry point into the system. It simplifies DeepMirror's extensive parameter space into a minimal, curated interface designed for rapid atmospheric exploration — particularly suited for non-specialised audiences.

image1.png

What the user sees

The Basic panel presents four main sections:

1. Carregar Exemplos (Load Examples) A single button that opens the System Images gallery, allowing the user to browse and select from previously generated atmospheres. Each image carries its full parametric preset; clicking it instantly reconfigures the entire system — prompts, sliders, menus — without requiring the user to understand any of these parameters.

2. Autoplay A Play/Stop toggle that automatically cycles through available presets at a user-defined interval (5–30 seconds). When activated, the system randomly selects presets from the loaded gallery, applying each one in sequence and sending all corresponding OSC values to TouchDesigner. This creates a continuous slideshow of atmospheric variations — a kind of automated "dreaming" that can inspire unexpected spatial ideas. A thumbnail and preset name update in real time.

3. Prompt A and Prompt B (read-only) Two read-only text areas that display synthesized summaries of the prompts operating in the Advanced mode. Prompt A is a Gemini-generated synthesis of Prompts 1 and 2; Prompt B synthesizes Prompts 3 and 4. These summaries update automatically (via debounced Gemini calls) whenever the underlying prompts change, giving the Basic user a readable overview of what the AI is generating without exposing the full prompt machinery. Below each prompt, a progress bar shows the averaged weight of the corresponding prompt pair.

4. Controles — Three Simplified Sliders Rather than exposing all twelve Advanced sliders, Basic Mode offers three intuitively named parameters:

Intensidade da IA (AI Intensity) — maps to the prompt-step-1 parameter, controlling how strongly the AI intervenes in the generation

Profundidade (Depth) — maps to the controlnet parameter, controlling the influence of the depth map on spatial structure

Retilíneo ↔ Orgânico (Rectilinear ↔ Organic) — maps to the black-level parameter, modulating the tonal quality of the generated atmosphere

A preview image displays the currently loaded preset's associated photograph.

Bidirectional Synchronization

A critical architectural feature of the two modes is their bidirectional nature: Basic and Advanced are not independent systems but two synchronized views of the same underlying state. This bidirectionality operates at multiple levels:

Basic → Advanced: When the user moves a Basic slider (e.g., "Intensidade da IA"), the corresponding Advanced slider (prompt-step-1) updates immediately, its value display refreshes, and the OSC message is sent to TouchDesigner — exactly as if the user had moved the Advanced slider directly.

Advanced → Basic: When the Advanced mode modifies a slider — whether through direct user interaction, preset loading, or gallery selection — the Basic slider and its percentage display update automatically via MutationObservers that watch the Advanced value display elements. The prompt synthesis areas also update: any change to the four Advanced prompt textareas triggers a debounced Gemini API call that regenerates the simplified summaries visible in Basic Mode.

Preset Loading → Both Modes: When a preset is loaded (from gallery click, Autoplay, or LOAD analysis), all Advanced parameters are set, which in turn propagates to all Basic displays. The Basic Mode's image preview, thumbnail, preset name, prompt summaries, slider positions, and weight bars all update to reflect the newly loaded state.

This bidirectional design means that a user can begin exploring in Basic Mode, switch to Advanced to fine-tune a specific parameter, and return to Basic without any discontinuity. The two modes are not hierarchical (simple vs. complex) but rather complementary views of the same creative state — one optimized for intuitive exploration, the other for precise parametric control.

Advanced Mode — Full Parametric Control

image2.png

The Advanced Mode exposes the complete DeepMirror II interface as documented in the preceding sections of this document. It provides:

  • 4-prompt architecture with column-based layout (Prompts 1 & 2 on the left, Prompts 3 & 4 on the right), each with independent Compose, Send LLM, and Send buttons
  • 10 atmospheric menus per column (5 fixed + 5 Gemini-generated) for combinatorial prompt construction
  • 12 real-time OSC sliders with direct numerical readouts, organized in functional groups (Prompt weights/steps, Image parameters, Depth map parameters)

SEED control with display field, range selection (1M–9M), and preset storage

Noise type selector (Perlin 3D, Harmonic, Alligator, Sparse) and Color/P&B toggle

Polarity control (Positive/Negative) affecting prompt interpretation

IA ON/OFF master switch

RECORD and LOAD workflows with full preset management

Gallery system with System Images, User Images, and custom folder browsing

Nano Banana architectural view generation module

Autoplay (▶ AUTO button in the title bar) cycling through presets with all Advanced controls visible

image3.png

The Advanced Mode is where the full depth of DeepMirror II's operative atmosphere design methodology is accessible — the complete set of parameters that define the chrono-operative practice of architectural atmosphere exploration. Although it presents numerous parameters—something that might initially intimidate first-time users—it ultimately proves to be highly accessible and explorable, even for beginners. Because each interface element has a perceptible impact on the variation of the generated image, users are encouraged to experiment with these parameters, thereby developing a dynamic and intensive interaction with the machine.