Skip to main content

Episode Generation

How AutoPod turns raw audio segments into polished, tagged episodes ready for distribution.

The generation pipeline

When a scheduled show finishes airing, AutoPod creates an episode through this pipeline:

1. Scheduled job

AutoPod's job generator runs every 5 minutes, scanning the broadcast schedule for shows that have recently finished. When it finds one, it creates a scheduled job — an instruction that says "create an episode from stream X, between time A and time B, for show Y."

2. Download segments

The episode generator reads the scheduled job and queries the segment index for all audio segments recorded between the start and end times. These segments are downloaded from cloud storage to a temporary workspace.

3. Stitch and encode

The downloaded segments are concatenated in chronological order and encoded into a single MP3 file using FFmpeg. If a pre-roll or sponsor tag is configured, it's prepended at this stage.

4. Tag and upload

MP3 metadata tags are embedded in the file:

  • Title — the show name and air date
  • Artist — the presenter name
  • Album — the show name
  • Year — the air year
  • Artwork — the show's cover image

The finished MP3 is uploaded to cloud storage and an episode record is created in the database.

5. Quality check

The episode checker validates the finished episode:

  • Is the duration within expected bounds? (A 1-hour show should produce roughly a 1-hour episode)
  • Is the file accessible in storage?
  • Are the metadata tags present?

If the check passes, the episode is marked as complete and becomes available to listeners. If it fails, the system can retry automatically.

6. Distribution

Once verified, the episode is immediately available through the Audio Player and Station Website. RSS feeds are regenerated every 2 minutes, so podcast subscribers receive the new episode shortly after.

Processing time

The time from a show finishing to the episode appearing depends primarily on the show's duration:

Show DurationApproximate Processing Time
15 minutes~1 minute
1 hour~2-3 minutes
2 hours~4-5 minutes
3 hours~6-8 minutes

AutoPod V2 is approximately 4x faster than V1 thanks to a parallelised processing pipeline.

Automatic vs manual episodes

The pipeline above describes automatic episode generation — triggered by the schedule. But you can also create episodes manually:

  • Manual upload — Upload an MP3 file directly through the dashboard. The file skips the recording and stitching stages and goes straight to metadata tagging and distribution.
  • Retroactive creation — Use the 60-day archive to create an episode from a past broadcast that wasn't in the schedule. You specify the time window, and AutoPod runs the same stitching pipeline.

All three methods produce the same result: a tagged, distributed episode in your catalogue.

What can go wrong?

Most of the time, episode generation works silently and perfectly. But occasionally things go wrong:

IssueWhat happensAutoPod's response
Missing segmentsStream was down during part of the showEpisode is created with shorter duration; gap is absent (not silent)
Generator failureProcessing error during stitchingAutomatic retry (up to 3 attempts)
Duration mismatchEpisode is significantly shorter than scheduledFlagged for review in the dashboard

You can see failed or retried episodes in the Episodes section of the dashboard, filtered by status.

tip

If an episode fails repeatedly, it usually means the audio segments are missing (stream was down). Check the 60-day archive — if the segments exist there, you can create the episode manually.