Get a Zoom meeting transcript with an API

#!/usr/bin/env bash
# Turn a Zoom cloud recording into a transcript with speaker names.
#
# Zoom's own transcript (the TRANSCRIPT file) exists only when audio transcription
# is on; when Zoom made one, it is saved as zoom-transcript.vtt. The audio file
# (M4A, or the MP4 when there is no M4A) is downloaded here, with your Zoom token,
# then sent to MeetWave for a transcript with speakers and their names.
#
#   export ZOOM_ACCOUNT_ID=... ZOOM_CLIENT_ID=... ZOOM_CLIENT_SECRET=...   # Server-to-Server OAuth app
#   export MEETWAVE_API_KEY=mwk_...
#   bash zoom-recording.sh <meeting id>
#   bash zoom-recording.sh zoom-audio.m4a      # an audio file you already downloaded
#
# Needs curl and jq.
set -euo pipefail

API_BASE="${MEETWAVE_API_BASE:-https://api.meetwave.io}"
API_BASE="${API_BASE%/}"

if [ -z "${MEETWAVE_API_KEY:-}" ]; then
  echo "Set MEETWAVE_API_KEY first. Get a key with one free hour: https://meetwave.io/speech-to-text-api?utm_source=github&utm_medium=integration&utm_campaign=api" >&2
  exit 1
fi
if [ "$#" -ne 1 ]; then
  echo "Usage: bash zoom-recording.sh <meeting id | audio file>" >&2
  exit 1
fi

# What to do when the API refuses the key or the balance is empty.
hint() {
  case "$1" in
    401) echo "Check MEETWAVE_API_KEY, or get a key: https://meetwave.io/speech-to-text-api?utm_source=github&utm_medium=integration&utm_campaign=api" >&2 ;;
    402) echo "Top up your balance: https://app.meetwave.io/developers?utm_source=github&utm_medium=integration&utm_campaign=api" >&2 ;;
  esac
}

# Calls an API; prints the answer, or prints the error answer and stops.
call() {
  local body status
  body=$(curl -s -w '\n%{http_code}' "$@")
  status="${body##*$'\n'}"
  body="${body%$'\n'*}"
  if [ "$status" -ge 400 ]; then
    echo "$status $body" >&2
    if [ -n "${MEETWAVE_CALL:-}" ]; then hint "$status"; fi
    exit 1
  fi
  echo "$body"
}
meetwave() { MEETWAVE_CALL=1 call -A "meetwave-examples/zoom" -H "Authorization: Bearer $MEETWAVE_API_KEY" "$@"; }

if [ -f "$1" ]; then
  AUDIO="$1"
else
  # 1. A Server-to-Server OAuth token, valid for one hour; list the recording's files
  ZOOM_TOKEN=$(call -X POST https://zoom.us/oauth/token -u "$ZOOM_CLIENT_ID:$ZOOM_CLIENT_SECRET" \
    -d grant_type=account_credentials -d "account_id=$ZOOM_ACCOUNT_ID" | jq -r .access_token)
  FILES=$(call "https://api.zoom.us/v2/meetings/$1/recordings" -H "Authorization: Bearer $ZOOM_TOKEN" | jq .recording_files)

  # 2. Zoom made a transcript: save it as it is
  TRANSCRIPT_URL=$(echo "$FILES" | jq -r 'map(select(.file_type == "TRANSCRIPT"))[0].download_url // empty')
  if [ -n "$TRANSCRIPT_URL" ]; then
    curl -sfL -H "Authorization: Bearer $ZOOM_TOKEN" -o zoom-transcript.vtt "$TRANSCRIPT_URL"
    echo "Zoom's transcript saved to zoom-transcript.vtt"
  fi

  # 3. Download the audio on your side, with your Zoom token
  FILE=$(echo "$FILES" | jq -c '(map(select(.file_type == "M4A")) + map(select(.file_type == "MP4")))[0] // empty')
  [ -n "$FILE" ] || { echo "This recording has no M4A or MP4 file." >&2; exit 1; }
  AUDIO="zoom-audio.$(echo "$FILE" | jq -r '.file_extension | ascii_downcase')"
  curl -sfL -H "Authorization: Bearer $ZOOM_TOKEN" -o "$AUDIO" "$(echo "$FILE" | jq -r .download_url)"
fi

# 4. Upload it to MeetWave and transcribe it with speakers and their names
UPLOAD=$(meetwave -X POST "$API_BASE/v1/uploads")
curl -sf -T "$AUDIO" "$(echo "$UPLOAD" | jq -r .upload_url)" || { echo "Upload failed" >&2; exit 1; }
ID=$(meetwave "$API_BASE/v1/transcriptions" -H "Content-Type: application/json" \
  -d "$(jq -n --arg id "$(echo "$UPLOAD" | jq -r .id)" '{upload_id: $id, speakers: true, speaker_names: true}')" | jq -r .id)

# 5. Wait while it is queued or processing, then print it
while :; do
  TRANSCRIPTION=$(meetwave "$API_BASE/v1/transcriptions/$ID")
  case "$(echo "$TRANSCRIPTION" | jq -r .status)" in
    queued|processing) sleep 5 ;;
    completed) break ;;
    *) echo "Transcription failed: $(echo "$TRANSCRIPTION" | jq -c .error)" >&2; exit 1 ;;
  esac
done
echo "$TRANSCRIPTION" | jq -r '.transcript.segments[]
  | "[\(.start | floor / 60 | floor | tostring | if length < 2 then "0" + . else . end):\(.start | floor % 60 | tostring | if length < 2 then "0" + . else . end)] \(.speaker // "Speaker"): \(.text)"'
echo
echo "$TRANSCRIPTION" | jq -r '"\(.id): \(.billing.audio_seconds) s of audio, $\(.billing.cost_usd)"'