Changelog

All notable changes to modelBridge are documented here.

v0.17 — Schema Resilience & Browse Polish (April 2026)

Added

“Coming soon” model cards — models that are live on fal.ai but not yet fully supported in modelBridge now appear as dimmed cards with a “Coming soon” badge instead of being hidden. modelBridge checks daily and automatically makes them available when ready — no plugin update needed. Learn more
Automatic schema recovery — if a model’s cached data becomes outdated after a plugin update, modelBridge now detects the issue and fetches fresh data from fal.ai automatically. Previously, affected models could appear with a blank form until manually re-added
Category sub-filter for Latest models — the “Last 14 days” filter now includes a category row so you can narrow down recent models by type (text-to-video, image-to-image, etc.)

Fixed

Error messages now link to help docs — when a generation fails, the error banner includes a “Learn more” link to relevant documentation with step-by-step fixes
Failed generations show correct status — the Billing tab now shows a clear “Failed” badge for generations that didn’t complete, instead of showing “No price” with no explanation
Error step instructions corrected — fixed cases where error messages referenced incorrect tab names or gave wrong instructions for clip duration adjustments

v0.16 — Model Intelligence & Academy (April 2026)

Added

Model preview system — expand any model card to see example output (image, video, or audio) before generating. Preview media extracted from fal.ai OpenAPI schemas with ~89% coverage across the catalog
AI-written key strengths — every model card shows 2–3 bullets describing what the model does well, written specifically for video editors via an automated Claude API enrichment pipeline
Technical specs row — resolution, duration, format, and aspect ratio information displayed directly on model cards
New model detection — manual refresh button for instant discovery, plus automatic background monitoring. “Latest models” filter shows models published within the last 14 days
modelBridge Academy links — contextual “Learn about X” links appear on model cards when a relevant Academy article matches the model’s capabilities
Show learning tips toggle — control whether Academy links and learning tips appear throughout the plugin. Set during onboarding, adjustable in Settings
Trial and subscription notifications — proactive alerts for upcoming trial expiry, subscription renewal, and payment issues

Performance

Parallel schema fetching — model schemas load concurrently for faster catalog browsing
Memory optimization — LRU caching, lifecycle cleanup, and leak prevention for stable performance during long editing sessions
Onboarding redesign — streamlined first-run experience with learning preference selection

v0.15 — Settings & Feedback (March 2026)

Added

Built-in feedback form — share ideas or report bugs directly from the plugin with screenshot attachments and automatic system context
Completion sound toggle — control the audio notification when background generations finish
Redesigned API key card — cleaner status display with “Bridge Connected” indicator

v0.14 — Full Catalog Validation (March 2026)

Added

1,229 models validated — every model in the fal.ai catalog scanned through the complete schema pipeline with zero errors
Automatic compatibility checks — each model is audited on install with results shown as warning badges, so you know before you generate
Pre-generation confirmation — models with known compatibility issues trigger a heads-up before spending API credits

Quality

1,082 of 1,082 audited models passed — 0 pipeline errors, 94.7% fully clean
148 automated tests across 6 suites

v0.13 — Category Gating (March 2026)

Added

11 supported generation categories — text-to-video, image-to-video, text-to-image, video-to-video, text-to-speech, and more — each with a dedicated pipeline
Automatic safety gating — unsupported or experimental model categories are blocked before they can cause failed generations
Audio pipeline — dedicated audio track import with intelligent track selection

v0.12 — Follow Your Generation (March 2026)

Added

Active Generations Panel — track all background generations with live status, elapsed timer, and progress steps
5-step progress tracking — Sent → Queued → Generating → Downloading → Importing with real-time updates
Snackbar notifications — floating success/failure toasts so you know when results are ready without checking the panel

v0.11 — Pricing Intelligence & Smart Import (March 2026)

Added

Live cost estimates — accurate costs before you generate, updated in real-time as you change parameters. The system checks multiple pricing sources and learns from your Metered cost history over time.
Cost labels — Metered (anchored to provider-reported usage units), Calculated (modelBridge formula), Estimated (curated rates), Learned (from your own past runs), From (starting price), and No price (honest “unavailable” when there is no rate to apply). See Costs & Pricing for what each label means today
Smart timeline import — one button, automatic routing: source-based models replace in-place, text-based insert at playhead, audio goes to audio tracks
Audio preview — play TTS and sound effect results directly in the preview card before importing

v0.10 — Self-Learning Validation (March 2026)

Added

Self-learning error prevention — when a generation fails due to input constraints, the plugin remembers and blocks the same mistake next time. The system gets smarter with every generation
Intelligent error handling — errors are caught before they cost money, learned from when they slip through, and shown in plain language with a clear next step — raw API errors are translated before they reach you
Adjacent clip detection — real-time detection for interpolation models, with direct two-clip timeline replacement
Server resilience — automatic crash recovery, health checks, and heartbeat monitoring

v0.9 — Dual Mode & Media Tools (February 2026)

Added

Dual Mode — run two models on the same clip simultaneously to compare results side by side
Dual-frame selection — models accepting start + end frames get a dedicated dual-column media card
Drag-and-drop from Finder — drop files directly onto media cards
Mask editor — built-in canvas editor for inpainting with adjustable brush, eraser, zoom and pan
Source Monitor preview — review results with In/Out point selection before committing
Prompt optimization — one-click AI rewrite of prompts for better generation results
“New” and “Commercial” filter chips — find recently launched models and commercially licensed models instantly

v0.8 — Cost Tracking & Reporting (February 2026)

Added

Real-time cost tracking per project and client with CSV export
Multi-currency support — USD, EUR, GBP, SEK, NOK, DKK, JPY, CAD, AUD with live conversion
HTML report generator — self-contained offline reports with charts, KPIs, and cost breakdowns by model and deliverable
External AI cost import — track spending from Midjourney, RunwayML, and other platforms alongside modelBridge costs

v0.5 — Model Discovery (January 2026)

Added

Model search — browse and add from fal.ai’s full catalog without leaving Premiere Pro
Built-in mask editor for inpainting workflows
Schema caching — models load instantly after first fetch

v0.3 — Schema-Driven UI (November 2025)

Added

Dynamic form generation — the plugin reads each model’s API specification and builds the interface automatically. 500+ models, zero hardcoded interfaces
Two-layer input validation — prevents wasted spend by catching errors before generation

v0.2 — Timeline Integration (October 2025)

Added

Auto-import to timeline — results land directly on the timeline, auto-positioned and auto-scaled
Node.js backend for media extraction and API communication
ExtendScript bridge for Premiere Pro timeline manipulation

v0.1 — Proof of Concept (October 2025)

Added

First working prototype — single model, basic CEP panel inside Premiere Pro
End-to-end generation: prompt → API → download → import