$ en/release-video-generator

Release Video Generator

Python pipeline that transforms a Jira HTML export into a narrated release video, with thematic script generation, text-to-speech synthesis, and automated video rendering.

Stack

Python OpenAI API BeautifulSoup FFmpeg Pydub

Project links

GitHub

Release Video Generator from Jira Tickets

This project automates the creation of a product update video from a Jira HTML export.

The idea is to turn technical tickets into a short, readable narrative format through a complete 5-step pipeline:

  • parsing Jira tickets,
  • thematic grouping and script generation,
  • text-to-speech generation for each segment,
  • audio normalization and assembly,
  • slide rendering and final MP4 export.

Context & Goals

Context

Release notes are often too long, inconsistent, and difficult to share with non-technical teams.

This project addresses that need by automatically transforming a Jira export into a narrated video that is easier to consume for:

  • product teams,
  • developers,
  • internal stakeholders,
  • and potentially, in the future, end users.

Goals

  • Automatically structure Jira tickets into coherent themes.
  • Generate a clear, accurate presentation script suitable for both screen text and voice-over.
  • Produce segmented audio narration synchronized with slides.
  • Generate a standardized final video (output/release.mp4) with no manual editing.

Pipeline Overview

flowchart TD A["Jira Export (HTML)"] --> B["Parse tickets (BeautifulSoup)"] B --> C["Thematic grouping + script (OpenAI)"] C --> D["TTS generation by segment (OpenAI)"] D --> E["WAV normalization + pauses + audio concat (FFmpeg)"] E --> F["Slide rendering + video concat + audio/video mux (FFmpeg)"] F --> G["release.mp4"]

Input Data and Artifacts

Element Description
Main input data/jira_export.html
Structured data build/release_structured.json
Generated themes build/themes.json
Video script build/script.json
Segmented audio audio/*.mp3
Final audio build/final_audio.wav
Timeline build/timeline.json
Final video output/release.mp4

Component Details

1) Jira Parsing (src/parse_jira.py)

  • Reads the Jira HTML export using BeautifulSoup.
  • Extracts key columns (type, key, summary, created, assignee, reporter, description).
  • Currently filters tickets by key prefix (OPD-).
  • Exports structured JSON for downstream steps.

2) Theme and Script Generation (src/generate_script.py)

The module runs 2 passes through the OpenAI API:

  • Pass 1: group tickets into functional themes (classification: UI, API, or unspecified).
  • Pass 2: generate final slides (intro, thematic slides, conclusion) with:
    • a title,
    • 3 to 5 on-screen bullet points,
    • a 3 to 5 sentence voice-over.

The requested tone is technical, clear, and non-marketing.

3) Audio Generation (src/generate_audio.py)

  • Reads build/script.json.
  • Generates one MP3 file per segment via gpt-4o-mini-tts (voice: alloy).
  • Names files in sequence (00_intro.mp3, etc.).

4) Final Audio Track Build (src/build_audio_track.py)

  • Converts MP3 files into normalized WAV (48kHz, stereo, PCM s16le).
  • Inserts silent pauses between segments.
  • Concatenates everything into build/final_audio.wav.

5) Video Rendering (src/render_video.py)

  • Generates one silent video slide per segment by overlaying text and panels on a background.
  • Adjusts each slide duration to match the real audio duration (via ffprobe).
  • Inserts visual pause segments.
  • Concatenates video segments, then muxes the final audio track.
  • Produces output/release.mp4 and a timeline JSON.

Full Orchestration (run_all.py)

The run_all.py script automatically chains all 5 steps together with error handling and per-phase logging.

Technical Stack

  • Language: Python 3.10+
  • AI (LLM + TTS): OpenAI API
  • HTML Parsing: BeautifulSoup4
  • Audio: Pydub + FFmpeg
  • Video: FFmpeg (drawtext, concat, mux)
  • Config: python-dotenv for API key management

Strengths

  • End-to-end, automated, reproducible pipeline.
  • Clear separation of responsibilities (dedicated script per step).
  • Clean artifact structure (build/, audio/, output/).
  • Robust audio/video synchronization through WAV normalization and duration probing.

Current Limitations

  • Strong dependency on the quality of the Jira HTML export.
  • Intentionally simple visual layout (static background, no advanced animations).
  • OpenAI models and parameters are hardcoded (limited external configuration).
  • Ticket filtering currently tied to the OPD- prefix.

Future Improvements

  • Configuration file support (models, voices, slide styles, Jira filters).
  • Multiple visual templates and rendering themes.
  • Automatic subtitle generation.
  • Internationalization (configurable narration language).
  • CLI packaging for “one command” CI/CD release workflows.

Related posts

No related posts yet.