TLDR

• Magic Mirror is live: capture a selfie, train, pick 3 looks, get results, animate. Fast, guided flow with clear progress and sensible defaults.

• Large template library with gendered variants and strong negative prompts for reliable portraits. Preview, generate more, then share or download.

• Quality of life: camera device switching, gallery fallback, private by default for images and videos, smoother media viewer, cleaner sharing links.

• Marketing and ops: QR promo code batches and print sheets, Magic Mirror landing at the edge, SendGrid sync, Tawk chat with user context, SEO updates.

Meet Magic Mirror

Magic Mirror is a guided portrait flow that turns a quick selfie into styled images using gendered templates and your subject details. It streamlines capture, training, look selection, generation, and optional animation. It runs on our custom ai models stack and a deep template system for sharp, consistent results, whether you use it as an ai headshot generator or a personal ai art playground. Try the new magic mirror mode: app.fiddl.art/magicMirror

The end-to-end flow

1. Capture

   • Live camera with device switching on desktop and mobile, mirrored preview, a louder shutter click, and a subtle flash so you know a frame was taken.

   • Graceful handling of permission issues with a clean fallback to gallery uploads. You can upload up to 10 selfies at once.

   • Higher preferred capture resolution and reliable inline playback on iOS for a smoother experience.

2. Train

   • We store subject gender and an optional subject description in your training set. That context is threaded into prompts for better fidelity.

   • Clear progress text and bars. Status states now include “starting” for more accurate feedback.

3. Pick looks

   • A dedicated dialog guides you to choose exactly three templates. The grid adapts to your screen and shows gender-appropriate previews.

   • Sticky confirm bar and loading skeletons keep the UI responsive while templates load.

4. Generate

   • Batch async creation submits your three looks together with consistent parameters, including negative prompts and your subject description.

   • New results are private by default. You can opt into public later if you want visibility.

5. Results

   • Mosaic gallery on desktop and a streamlined list on mobile. Sticky actions for Start again, More looks, and Forge.

   • Per-image actions: share, download, animate. We use reliable blob downloads and the native Share API where available.

   • “Generate more for template” schedules extra renders per look. We track pending states and poll so new images appear as soon as they are ready.

6. Animate

   • Turn favorite stills into short videos. We estimate credits by duration and model pricing before you commit.

   • One-click guard prevents retriggering animation on the same image. Videos default to private and 9:16 for vertical viewing.

Templates that stay on model

Magic Mirror includes a large, gendered template library designed for portrait consistency and style range. Each template defines camera notes, composition, lighting, color, mood, and post-processing, plus robust negative prompts to avoid close crops, occlusions, and artifacts. Highlights include:

• Everyday and editorial: LinkedIn-ready business headshots, dating profile portraits, streetwear at night with cinematic neon, fitness-in-gym looks.

• Culture and genre: Red Carpet Event, Cyberpunk Neon Portrait, Spacefleet Officer, Wizard House Portrait, K-Pop Idol.

• Lifestyle sets: Sunset Beach Traveler, Cozy Café Creator, Festival Glow, Outdoor Adventurer.

Each look has male and female variants with tailored wardrobe and pose guidance. After you see results, browse more looks or schedule extra renders per template. It lets you make ai art with custom models without constant prompt tweaks.

Under the hood

• Template APIs: Filter by tags and kind, fetch gendered previews, and use consistent slots. OpenAPI docs are updated for template listing and batch creation.

• Batch async creation: One call schedules multiple images with acceptance feedback and a batch ID. Status values include starting, pending, processing, succeeded, and failed.

• Training set metadata: New fields for subject gender, subject category, and a free-text description. A detail endpoint supports richer curation and search.

• Model support and quality: Wider model list for images and video. LoRA scale nudged from 0.95 to 0.98 for stronger customizations in Flux-based models.

• Privacy by default: New image generations and animations are private unless you set them public. Video animations default to 9:16.

• Shareable request links: Short base64url IDs replace long UUIDs in links. We validate short IDs on the edge for safety.

• Edge safety and logging: Centralized wrappers standardize error handling and structured logs across edge routes, with graceful fallbacks for reliability.

Camera and gallery polish

• Live device switching with persistence, clearer errors, and stable inline playback on iOS.

• Quicker gallery selection with a clean file picker. You can re-select the same file without a refresh.

• Media viewer fixes: no accidental closes when tapping creator info, better HD load timing, and a resolved infinite-loading case.

Marketing and ops

Promo codes, QR, and print sheets

• QR code batch generator: Create hundreds of unique claim codes with manifests in CSV and JSON for tracking.

• A4 sheet builder: Automatic 3 x 3 layouts at 300 DPI with label parsing, sorting, and an output manifest for audits.

• Promo claim UX: URL detection, short-to-long code conversion, one-time claim handling, auth-aware flows, and clear notifications.

• Admin shortcuts: Copy a standard claim link or a Magic Mirror link that embeds the code.

• I showed up to Nas Summit in Bangkok with 100 trial codes and introduced the platform to dozens of creators.

Growth and comms

• Magic Mirror landing at the edge: SEO-friendly HTML, Open Graph, JSON-LD, and fast delivery.

• Email delivery: All user emails sync to SendGrid for reliable transactional and marketing sends.

• Live chat: Tawk integration with user ID, points, and avatar metadata so support has context.

• SEO updates: Robots and sitemap tuned for key pages. A crawler guide helps LLM bots understand site structure.

Other improvements

• Use your uploads for video: Upload custom start images with presigned URLs. Browse your uploads from a new gallery picker.

• Veo3 image input support: You can now feed a starting image into Veo3 requests. Choose 720p or 1080p output.

• Model discovery: Filter models on our Models Page by author, faster single-model page loading, and the public models endpoint is open without auth so anyone can browse our available models.

• Profile revamp: Unified profile and account, clearer creation capabilities, and a polished model and feature overview.

• Pricing clarity: Points pricing updated across models, with cost estimation shown for animations.

Who is this for

Creators who want fast, consistent portraits without prompt fatigue, teams running events that need scannable claims and clean print assets, and anyone exploring text to image ai with a curated, reliable path. If you use Forge, the same foundation works as a personal face forge generator for style experiments in your custom ai models.

Notes and small fixes

• Stronger error handling across async flows, with explicit exit codes for script failures.

• Creation stores dedupe identical video queries to avoid duplicate loads.

• Smoother login prompts in Magic Mirror and better redirects back to where you started.

• Typography and layout refinements for galleries, template grids, and sticky action bars.

Getting started

1. Open Magic Mirror and capture a selfie or upload a few.

2. Add a short subject description if you want extra guidance.

3. Pick three looks and confirm. Watch training and generation progress.

4. Download, share, generate more, or animate your favorites.

Remember, you can always follow our progress on Trello.