Skip to content

v0.1.0 - Multi-LLM Moment Analysis, Selectable Caption Styles & Core Pipeline Fixes

Choose a tag to compare

View all tags
@JayWebtech JayWebtech released this 23 Jun 07:50
· 5 commits to main since this release
v0.1.0-beta
c4c3a3b

AutoShorts v0.1.0 Release Notes

We are excited to release the foundational build of AutoShorts! This build provides local-first portrait video cropping, AI-powered viral moment analysis, and fully local caption burn-in styling.

🚀 Key Features & Updates

1. Interactive Caption Style Selector Modal

When importing a recording, the application now displays an interactive overlay Modal allowing you to select how automated captions should look on generated portrait shorts.

2. 7 Modern & Premium Design Presets

We have added seven distinct captions presets, rendered locally in FFmpeg using high-end system typography (prioritizing Futura and Avenir Next collections) and dynamic padding:

  • Modern Box (Default): Cinematic uppercase white text inside a semi-transparent black padding box.
  • Classic Outline: High-energy yellow bold text with a thick black outline (iconic CapCut/TikTok formatting).
  • Minimal Shadow: Pure white text with a soft drop shadow.
  • Vibrant Cyan: Modern tech-themed cyan text with a drop shadow.
  • Vibrant Yellow Box: Bold black text inside a solid yellow padding box (Motivational/Reels style).
  • Vibrant Green: High-contrast neon green text with outlines and shadows (Alex Hormozi style).
  • Vibrant Red: Bold neon crimson text with drop shadows (Gaming/dramatic style).

3. Responsive Font & Layout Scaling

Captions are no longer static. The application automatically probes video dimensions and scales the font size and padding box proportionally based on the cropped portrait width. This guarantees text fits perfectly on any output resolution (from 202px width up to 1080px).

4. Absolute Timeline Synchronization

Fixed a bug where captions were missing/skipped on outputs. Shipped absolute timestamps alignment in the FFmpeg filter graph to match output-seeking Presentation Timestamps (PTS).

5. Robust Fallback Rendering

Improved error handlers to rescue from captioned render failures. If captioned rendering fails for any reason (e.g. missing fonts, local character escape conflicts), the application automatically falls back to rendering a clean crop without captions, saving the detailed error logs to the candidate’s DB entry.

6. Dynamic Multi-LLM Dispatcher (DeepSeek & Claude)

Supports both DeepSeek (default) and Claude (Anthropic) for moments detection. The active LLM settings and warnings dynamically adapt depending on the active LLM_PROVIDER environment variable, visible inside the collapsible API credentials sidebar.

7. Native Project Management

Added full database and backend support to Rename or Delete projects directly from the dashboard grid.

🛠️ Installation & Setup (FFmpeg Filter Requirements)

For captions to burn in locally, a full-featured build of FFmpeg is required.

  1. Install FFmpeg using Homebrew: 
    brew install ffmpeg
Assets 3
Loading