Repository navigation
Reeltern 0.5.0 · Local editing for people and AI
Pre-releaseReeltern 0.5.0 — AI Director, creator-grade polish, and closed-loop QA
Edit through Reeltern Studio (desktop), the command line, or 31 AgentFlight Grade-A MCP tools for Claude Code, Codex, Cursor, VS Code, and other local clients.
Downloads
- Windows:
reeltern-0.5.0-windows-python.zip - macOS (Apple Silicon or Intel):
reeltern-0.5.0-macos-python.zip - Linux:
reeltern-0.5.0-linux-python.zip - Existing Python environment:
reeltern-0.5.0-py3-none-any.whl - Source:
reeltern-0.5.0.tar.gz - Download verification:
SHA256SUMS.txt
These are Python installation bundles, not standalone native apps. Install Python first (3.12 recommended), extract the ZIP, and run python3 install.py on macOS/Linux or py -3.12 install.py on Windows. The installer creates a private .venv and installs dependencies locally without changing global packages. See the included START-HERE.md.
What is new in 0.5.0
- AI Director (
compose_video/reeltern compose --style ...): One-call production presets (cinematic,viral_reel,tutorial,documentary,noir_trailer) pairing silence trimming, intro slates, aspect framing, motion, color grading, voice cleanup, music ducking, loudness normalization, and captions. - Closed-loop render QA (
audit_render/reeltern audit JOB_ID): Measures EBU R128 integrated loudness (-16 LUFS), true peak, luminance, contrast, near-black frames, rendered contact sheet PNGs, and actionable repair fixes. - Creator-grade video & audio:
blur_fillvertical halo framing, Ken Burns camera motion (zoom_in,zoom_out,pan_left,pan_right), 10 color grades, 10xfadetransitions,kineticword-highlight burned ASS captions, sliding bottomprogress_bar,voice_cleanupEQ/compressor, andimport_audio+sidechaincompressmusic ducking. - MCP visual inspection & ergonomics: Expanded to 31 tools (
100/100 Grade A,0errors,0warnings), binary PNG resource templates (reeltern://visual/asset/{asset_id},reeltern://visual/render/{job_id}),reeltern://cookbook, andviral_short/polish_editprompts.
Release checks & limits
Publication is gated on core unit tests (53 checks locally, including pixel, audio, CLI, Studio Tk, and live stdio MCP checks) plus clean bundle installation, real demo render, cache-reuse checks, and 31-tool MCP handshake across Python 3.10–3.12 on Ubuntu, Windows, and macOS. Python and internet access are required for bundle dependency setup; GUI requires Tkinter. Bundles remain unsigned; there is still no built-in speech recognition or face tracking.
Showcase & Downloads · Install guide · AI setup · Validation