What's New

GRAPHIC — Computational Plate Studio

A full graphic-design mode with a live Plate Visualiser, ~48 industry-anchored presets, deterministic composition and a batch export pipeline (SVG/PNG/contact sheet).

• ▸Archetype component library across Dial, Circuit, GridMicro, Signal, Topo, Chart, TypeSpecimen and more.
• ▸Disciplined free-flow layout with proximity grouping, clear-zones, density guard and per-role size steppers.
• ▸Text toggles + SHUFFLE TEXT / SHUFFLE SIZE for controlled randomness.
• ▸Batch Preview Strip: see 6/12/24 tiles before export, reshuffle in place, click to peek.

Witness Copilot (Beta)

An admin-gated assistant that reads your current selection and prompt to analyse, compare, score confidence, and hand off a seed to a fresh Atelier session.

• ▸Selection bridge reads gallery metadata for real context.
• ▸Prompt Confidence Check scores five axes (subject, style, camera, light, composition).
• ▸Send-to-Atelier opens a brand-new session pre-filled with your seed.

Billing hardening & regional pricing

Signed credit ledger with reconciliation audit table, prorated plan upgrades (pay the delta), and INR/UPI shown first to Indian users on plans and top-ups.

• ▸credit_reconciliations audit trail — no silent backfills.
• ▸Prorated upgrade orders charge only the difference.
• ▸Geo-based currency defaults across pricing surfaces.
• ▸Admin ledger grants recorded in usage logs.

Mobile & tablet layout pass

Header collapses to icons on tablet, mode chips wrap without clipping, Flash opens as a modal with a real gallery, and Neural Link opens full-view on mobile without iOS auto-zoom.

Learn Mode & What's New

Contextual ? tooltips read straight from the Operator's Manual, and a headless What's New modal surfaces new changelog entries automatically — once per user, per entry.

Polymorphic Director Console

The Director now works in four modes — Automotive, Fashion, Character and Mecha — from one console. Build a shot axis by axis from labelled grids; each mode has its own deep vocabulary.

• ▸Fashion: casting (incl. heritage & skin tone), garment, beauty, accessories and styling — ~35 categories.
• ▸Character: archetype, race, class, wardrobe, weapons, powers and render style.
• ▸Mecha: chassis, systems, hull, greebles, weapons and wear.
• ▸Multi-select for accessories, weapons, greebles and effects; contextual filtering (gender to hair/makeup, house to palette).

The Gallery

Renders now collect in a visual grid you can revisit, compare and re-open. Multi-frame sets stay together as contact sheets; the Viewer opens any image next to the prompt that made it.

The Operator's Manual

A full in-app manual covering every engine, control, the parameter deck, the Inner Atelier and delivery — with playbooks and troubleshooting.

Reproducibility — Seed, Lock & Re-roll

Lock a seed to reproduce a look, lock the segments you like, and re-roll only the rest. A slot machine becomes a repeatable process.

The Inner Atelier

Three specialist masters — Autocraft, Lumen and Sigma — critique and hand off to build production-grade prompts, plus a full Poster Forge.

Flash rendering + model picker

Generate images inline and choose the model behind each render — fast models for iteration, higher-fidelity models for finals.

Ship anywhere — multi-target export

One prompt, formatted on demand for Midjourney, Gemini, SDXL or Flux — parameters and negatives handled per platform.

Plans, credits & billing

Four tiers (Explorer, Operator, Architect, Studio) with USD/INR pricing, credit rollover on Architect/Studio, and one-off top-ups.

Public site & landing page

Witness Labs is live with a full landing page and the workstation at /app.

Prompt engine quality pass

Cleaner prompts across every engine: removed dated 'quality-token' filler, tighter selection logic, auto-negative prompts, and fixed subject-burying.