Merge pull request #23 from soniasriv/add-gladia-podcast-shorts

Add gladia-podcast-shorts example

Author: dazzatronus

examples/gladia-podcast-shorts/.env.example

@@ -0,0 +1,21 @@
+# Your Shotstack API key: https://dashboard.shotstack.io
+SHOTSTACK_API_KEY=
+
+# Your Gladia API key: https://app.gladia.io
+GLADIA_API_KEY=
+
+# A public link to the recording you want to cut up
+AUDIO_URL=
+
+# Optional. Topics the clips should cover, separated by commas.
+KEYWORDS=
+
+# Optional. A public video to play behind the captions.
+BACKGROUND_URL=
+
+# Optional. How many clips, and how long each one can be.
+CLIP_COUNT=
+CLIP_SECONDS=
+
+# Optional. Set to stage for free watermarked test renders.
+SHOTSTACK_ENV=

examples/gladia-podcast-shorts/.gitignore

@@ -0,0 +1,2 @@
+.env
+output/

examples/gladia-podcast-shorts/README.md

@@ -0,0 +1,60 @@
+# Gladia podcast to vertical shorts
+
+Cut a long recording into short vertical clips. The script sends one audio URL
+to Gladia, reads the transcript with speaker labels and word timings, picks the
+segments that say the most, and renders each one as a 1080x1920 MP4 with
+captions and a speaker tag. You get the clips in `output/` and a manifest that
+records where each one starts in the original recording.
+
+The subtitles are cut from the transcript the script already has, so the render
+does not transcribe again and consumes no generation credits.
+
+## Requirements
+
+- A [Shotstack account](https://dashboard.shotstack.io/register) and your
+  **production** API key
+- A [Gladia account](https://app.gladia.io) and an API key
+- A public link to an audio file, one to three hours long
+- Node.js 20 or later
+
+Each run transcribes the recording once and renders one video per clip.
+Set `SHOTSTACK_ENV=stage` in `.env` and use your sandbox key for free
+watermarked test renders.
+
+## Setup
+
+```bash
+git clone https://github.com/shotstack/shotstack-cookbook.git
+cd shotstack-cookbook/examples/gladia-podcast-shorts
+cp .env.example .env
+```
+
+Load the file into your shell. Do this in each new terminal:
+
+```bash
+set -a
+source .env
+set +a
+```
+
+Set `AUDIO_URL` to your recording. Set `KEYWORDS` to the topics you want the
+clips to cover, separated by commas. Set `BACKGROUND_URL` to a public video if
+you want moving footage behind the captions.
+
+## Run
+
+```bash
+node podcast-shorts.mjs
+```
+
+## What happens
+
+The script asks Gladia to transcribe the recording with diarization, then polls
+every five seconds until the status is `done`. It builds every window of
+consecutive speech up to `CLIP_SECONDS` long, scores each window by how much is
+said in it, and keeps the highest scoring windows that do not overlap. For each
+one it uploads a subtitle file to the Shotstack Ingest API, renders a vertical
+video that trims the source audio to that window, and downloads the MP4.
+
+Run the script again and it reuses the transcript, so only the renders repeat.
+To change how many clips you get, set `CLIP_COUNT` in `.env`.