Skip to content

SubBuild Studio v2.0.0

Choose a tag to compare

@talevi83 talevi83 released this 18 Nov 22:34
· 39 commits to main since this release
8973141

πŸš€ SubBuild Studio v2.0.0 β€” Multi-Provider AI Engine, Smart Cost/Quality Optimizer & Bento Settings Overhaul

SubBuild Studio v2.0.0 is a major architectural milestone that transforms the studio from a single-provider tool into a full Multi-Provider AI Subtitle & Translation Workbench, supporting Google Gemini, OpenAI, Anthropic Claude, and Local / Cloud Ollama with intelligent hardware-aware model selection and a redesigned settings experience.


✨ What's New in v2.0.0

1. 🌐 Multi-Provider AI Architecture & Unified LLM Routing

  • 4 AI Providers Across Every Pipeline Stage: Mix and match Google Gemini, OpenAI, Anthropic Claude, and Ollama (Local GPU or Ollama Cloud) independently for:
    • Subtitle Translation
    • Series Glossary & Character Gender Discovery
    • AI Vision / On-Screen Text (Forced Narratives OCR), Speaker Detection & Sync Line Matching
  • Multi-Account API Key Vault (storage/provider_keys.json):
    • Store, label, test, enable/disable, and set default API keys per provider directly from the UI.
    • Supports custom base URLs and one-click Ollama Cloud preset configuration.
    • Secrets are masked in all API responses and never overwrite keys on view reloads.
  • Live Model Discovery & Capability Filtering:
    • Automatically queries live provider endpoints on startup and key refresh (GET /v1/models, Anthropic models API, Gemini models API, and Ollama /api/tags + /api/show).
    • Deduplicates dated OpenAI snapshots, verifies real Vision capability on local Ollama models via /api/show, and strictly filters dropdowns so each stage only shows models that are verified, available for your loaded keys, and capable of that specific task.

2. πŸ’‘ One-Click "Load Best Cost / Value Models" (GPU & VRAM Aware)

  • Smart Auto-Recommender in Settings: Automatically scans all configured API keys and selects the most cost-effective model that delivers top-tier quality for each pipeline stage.
  • Google Flex Tier Integration: Automatically factors in 50% Google Flex discounts when comparing cloud pricing across providers.
  • Hardware-Gated Local AI (Include Local AI):
    • Optional checkbox to include free local Ollama models ($0 token cost) in the optimization.
    • Inspects your installed GPU and available VRAM (e.g., 8 GB vs. 12 GB+ vs. CPU) and only selects local models that fit comfortably in VRAM and are proven for high-quality translation/vision β€” automatically skipping oversized models (14B / 20B+ on 8 GB GPUs) with a clear explanation badge.

3. 🎨 Settings Screen Deduplication & Compact Bento Grid Redesign

  • Eliminated Duplicate Controls: Removed job-specific subtitle toggles (Metric Conversion and Remove SDH Speaker Labels) from the global Settings screen so they live exclusively in the post-video Pipeline Configuration screen where per-video presets are chosen.
  • 2-Column Bento Layout:
    • Row 1: Appearance (Language & Theme) alongside a compact Local AI & Cost-Saver status card with a collapsible setup drawer.
    • Row 2: Streamlined API Key Vault with provider tabs, inline connection status, and quick-apply actions.
    • Row 3: Side-by-side 3-Column AI Stage Cards (Translation, Glossary, Vision) with per-episode (45m) and feature-film (120m) cost badges, suitability warnings, and a collapsible Interactive Cost Estimator.
    • Row 4: Storage Cleanup alongside Diagnostics & Feedback (Server Log download, Bug Report, and Feature Request shortcuts).

4. πŸ› οΈ Studio, Ingest & Localization Enhancements

  • Multi-Stage Status in Sidebar & Ingest: Summarizes active providers across all pipeline stages (e.g., OpenAI Β· Ollama) and warns accurately if any active stage is missing a configured key.
  • Row Re-Translation Dialog: Filters candidate models by configured keys and translation capability, and dynamically shows/hides Thinking levels based on the selected model's provider.
  • Full Bilingual Support (EN / HE): Complete RTL/LTR localization across all new multi-provider controls, benchmarks, tooltips, and Help & Q&A documentation.
  • Expanded Enterprise QA Suite: Verified with 504 Pytest backend tests, 13 Vitest frontend component tests, and 52 Java 21 (Selenium 4 + RestAssured) black-box E2E/API tests.