Portfolio
AI-Assisted Desktop Application

AskLibra Shorts

A local desktop workflow for turning long videos into short-form content through transcription, clip suggestions, face-aware reframing, caption styling, edit planning, preview rendering, and platform-ready exports.

Role

Product Designer, Desktop Application Architect, AI-Assisted Full-Stack Developer

Platform

Desktop Application

Status

Live

Year

2026

Private desktop application environment

This case study focuses on the product workflow, local media pipeline, interface behavior, and export process without exposing private video files, local paths, or unpublished content.

Overview

AskLibra Shorts is a desktop application designed to reduce the number of tools needed to repurpose long-form video. It combines local media import, audio extraction, transcription, clip analysis, reframing, caption styling, previewing, edit planning, and rendering inside one project-based workspace.

The workflow emphasizes local processing where practical, rather than presenting short-form video production as a cloud-only application.

Problem

Producing a polished short often requires several disconnected tools for transcription, clip selection, reframing, captions, previewing, and rendering.

  • Long-form footage requires manual review
  • Identifying strong short-form moments is time-consuming
  • Transcription often happens in a separate tool
  • Vertical reframing can lose the speaker's face
  • Captions require separate styling and rendering
  • Preview and final export may use different workflows
  • Creators move files between several applications
  • Render progress and output location can be unclear
  • Local projects can break when stored on disconnected drives
  • YouTube imports require additional download handling

Solution

AskLibra Shorts brings the most repetitive parts of short-form video production into one desktop workflow. The application analyzes local media, creates a transcript, recommends clip ranges, applies caption and reframing choices, generates an edit plan, and renders the selected result using local media tools.

  • Reduce tool switching
  • Make clip discovery faster
  • Preserve selected caption settings
  • Keep faces visible in vertical exports
  • Provide reliable preview and final rendering
  • Support multiple platform aspect ratios
  • Keep users informed during long-running operations
  • Maintain stable local project storage

Previous workflow

The process depended on several tools and repeated manual steps. This increased editing time and made it difficult to keep transcription, clip timing, caption styling, preview output, and final rendering synchronized.

Import footage into an editor -> extract or generate a transcript -> review the full video -> manually identify clips -> reframe for vertical video -> add and style captions -> render a preview -> make corrections -> export for each platform.

Core workflow

The desktop workflow moves from source import to local transcription, clip selection, edit planning, preview, and final export.

  1. 1

    Import video

    Add a local video file or import supported online video content.

  2. 2

    Prepare media

    Copy or register the source inside the project and extract audio for transcription.

  3. 3

    Transcribe locally

    Use whisper.cpp to generate a transcript and SRT caption data.

  4. 4

    Analyze transcript

    Review transcript segments and identify candidate moments for short-form clips.

  5. 5

    Suggest clips

    Present recommended clip ranges with timing and scoring information.

  6. 6

    Select and adjust

    Choose a suggested clip or manually refine its start and end points.

  7. 7

    Configure format

    Select 9:16, 1:1, or 16:9 output based on the destination platform.

  8. 8

    Apply reframing

    Reframe the selected segment while keeping the primary speaker visible.

  9. 9

    Customize captions

    Choose caption style, position, chunking, and visual treatment.

  10. 10

    Create edit plan

    Save timing, layout, caption, and reframing instructions as a reusable render plan.

  11. 11

    Preview

    Render a lower-cost preview so the user can inspect the result before final output.

  12. 12

    Render and export

    Produce the final short and save it inside the project's export folder.

Feature areas

The application groups the short-form production workflow into import, transcription, reframing, captions, planning, preview, and export.

Import and project setup

Creates a consistent project workspace for source media, transcripts, edit plans, previews, and final exports.

  • Local video import
  • Supported YouTube import
  • Project folders
  • Stable application-support storage
  • Source and export organization

Transcription and clip discovery

Converts long footage into searchable transcript data and surfaces candidate moments for short-form editing.

  • Local audio extraction
  • whisper.cpp transcription
  • SRT generation
  • Transcript segmentation
  • Suggested clips
  • Clip scoring
  • Timing adjustment

Reframing and format control

Adapts the selected clip for different social platforms while preserving the most important visual subject.

  • 9:16 export
  • 1:1 export
  • 16:9 export
  • Face-aware cropping
  • Platform-oriented framing
  • Manual format selection

Caption workflow

Keeps caption choices synchronized from the interface through preview and final rendering.

  • Caption style selection
  • Caption chunking
  • Words-per-caption control
  • Position and layout options
  • SRT-based burn-in
  • FFmpeg caption rendering
  • Selected-style persistence

Edit planning and preview

Separates planning, previewing, and final output so users can validate the edit before committing to the full render.

  • Edit-plan generation
  • Saved plan state
  • Preview rendering
  • Per-clip plan status
  • Reset plan
  • Final render enabled only when a plan exists

Export and feedback

Makes long-running media operations visible and helps users locate completed files without guessing.

  • Persistent export status
  • Progress feedback
  • Success and error states
  • Saved-file hints
  • Open Exports support
  • Modal and inspector feedback
  • Final output organization

My role

I defined the product workflow, desktop application structure, media-processing pipeline, interface direction, project-storage behavior, caption and reframing controls, preview strategy, and export feedback. AI-assisted development accelerated implementation, debugging, refactoring, and iteration, while workflow decisions, usability testing, and final validation remained under my direction.

Product DesignDesktop ArchitectureVideo Workflow DesignMedia Pipeline DesignUX DirectionAI-Assisted DevelopmentManual TestingRelease Preparation

AI-assisted, human-directed development

AI accelerated implementation and troubleshooting across the React, Rust, and media-processing layers. Product direction, workflow design, visual judgment, usability decisions, and final testing remained human-led.

AI accelerated

  • React and TypeScript implementation
  • Rust and Tauri integration
  • FFmpeg command construction
  • Filter-chain debugging
  • Caption-rendering fixes
  • File-path troubleshooting
  • UI iteration
  • Build error diagnosis
  • Documentation
  • Test planning

Human direction and ownership

  • Product requirements
  • Editing workflow
  • Interface design
  • Local-processing decisions
  • Media pipeline behavior
  • Caption requirements
  • Reframing expectations
  • Manual visual review
  • Workflow validation
  • Release decisions

Architecture and stack

Technology stack

TauriReactTypeScriptRustFFmpegwhisper.cppViteyt-dlpSRTLocal processing

Technical challenges

01

Coordinating multiple runtimes

Connected a React interface, Tauri commands, Rust application logic, and local media binaries inside one desktop workflow.

02

FFmpeg filter-chain reliability

Built and debugged media commands involving cropping, scaling, caption burn-in, audio handling, and multiple output formats.

03

Caption-style persistence

Resolved a bug where captions rendered correctly only when users left the default style unchanged.

04

Face-aware reframing

Applied dynamic crop behavior intended to keep speakers visible in vertical-video outputs.

05

Preview versus final rendering

Separated edit-plan creation, preview output, and final rendering so users could validate results before export.

06

Long-running operation feedback

Added persistent status messaging so transcription, planning, previewing, and exporting did not appear frozen.

07

Stable local project paths

Moved project storage away from a removable external drive into the application-support directory to prevent broken paths when the drive was disconnected.

08

Local and online imports

Managed differences between local files and downloaded video sources while keeping the downstream workflow consistent.

09

Smaller-window usability

Refined layout and overflow behavior so the editor remained usable near the application's minimum supported width.

10

Dependency and release readiness

Verified which local binaries and runtime dependencies must be present for the desktop application to function after installation.

Key fixes and iterations

Product development focused on making the desktop workflow more reliable across local files, downloaded sources, captions, preview, and export.

  • External-drive project path failures
  • YouTube filename mismatch
  • FFmpeg caption filter issues
  • Preview overflow on smaller screens
  • Selected caption styles not reaching final render
  • Persistent export-status feedback
  • Compact start and end clip controls
  • Edit-plan state and final-render gating
  • Application branding update
  • Project storage moved to Application Support

Quality assurance

Build integrity

  • TypeScript checking
  • Rust compilation
  • Tauri development builds
  • Production build checks
  • FFmpeg command validation
  • Dependency-path checks

Functional validation

  • Local import testing
  • Supported online import testing
  • Transcription testing
  • SRT generation
  • Suggested clip review
  • Timing adjustment
  • Caption-style testing
  • Face-aware reframing review
  • Preview rendering
  • Final rendering
  • Export-folder verification

UI and live validation

  • Minimum-window-width behavior
  • Side-panel layout
  • Modal behavior
  • Status feedback
  • Error-state visibility
  • Project reopening
  • Edit-plan persistence

Results

Qualitative workflow outcomes

One local editing workflow

AskLibra Shorts combines the main steps of video repurposing into one desktop workflow.

Local transcription and SRT generation

The application generates transcript and subtitle data for the selected source media.

Transcript-based clip suggestions

Transcript analysis helps surface candidate clip ranges while preserving user control.

Face-aware multi-format exports

Selected clips can be prepared for 9:16, 1:1, and 16:9 output with speaker-aware framing.

Configurable burned-in captions

Caption style, position, chunking, and rendering choices flow into preview and export.

Preview-before-render workflow

Users can inspect a lower-cost preview before committing to the final render.

Persistent export feedback

Long-running media operations show status, success, error, and saved-file feedback.

Project-based local storage

Project storage keeps source media, transcripts, edit plans, previews, and exports organized locally.

Project takeaways

AskLibra Shorts combines the main steps of video repurposing into one desktop workflow, reducing tool switching while preserving user control over timing, framing, captions, previewing, and final export.

Local AI can improve creative workflows

Transcription and analysis can run within a desktop application, reducing dependence on separate cloud tools for every stage.

Preview and final output should share one plan

Saving timing, caption, framing, and format decisions in an edit plan helps prevent preview and final-render mismatches.

Long-running operations need visible state

Transcription and video rendering may take time, so clear persistent feedback is part of the core product experience.

Media applications require path discipline

Stable project and binary paths are essential because removable drives and renamed downloads can break otherwise-correct workflows.

What this project demonstrates

AskLibra Shorts demonstrates how local AI, desktop application development, and media-processing tools can be combined into a practical creator workflow.

  • Desktop application architecture
  • Local AI transcription
  • FFmpeg media processing
  • Face-aware video reframing
  • Caption and subtitle workflows
  • Project-based file management
  • Human-directed AI development
  • Iterative product debugging