Internal company podcast: a weekly update made from a changelog

An internal company podcast made by API: six changelog entries became a weekly update for the team. Hear it, copy the request, and see what we found.

How this was made. Audio and video overviews, slide decks, infographics and quizzes are made with Google's Gemini Notebook (formerly NotebookLM), on accounts that AutoContent operates. AutoContent has no agreement with Google. Google can limit or stop this access at any time.

The episode

AutoContent API changelog, October and November 2026 (status.autocontentapi.com)

PodcastEnglish5 min 36 s

0:005:36

Transcript excerpt

Speech-to-text of the output, not edited and without speaker labels. It is part of the sample output.

And that exact philosophy, stopping the waste of time on doomed requests, that seems to be driving the video API updates too. It really is. Like, if video generation fails now, it fails immediately with a clear error message. Which is huge. Yeah. Previously, it would just hang in limbo for about 90 minutes before throwing a generic error.

And I'll quickly round up the final fixes here, too, because it looks like we're applying this logic across the board. Yeah, we are. On October 28th, we stopped partial download drops on data.autocontentapi.com. And on October 7th, we ensured long intro titles actually fit inside the video frame.

AI-generated with AutoContent APIDownload MP3Full transcript (TXT)
Length
5 min 36 s
Credits used
10 credits
Cost on the Starter plan
€0.29
Generation time
4 min 47 s
Format
Podcast
Job
Company podcasts
Language
English
Generated on
2026-11-02
Make this with the API

This output used 10 credits, which is €0.29 on the Starter plan (€29 per month for 1,000 credits). Failed generations use none. Generation time runs from the request to the finished file; it depends on the sources, the options and the queue.

What came out, and what we would change

We pasted the six entries of our own public changelog for October and November 2026 as text and asked for this week's episode of an internal podcast for the whole team: what changed for customers, why it matters, and what support should tell customers, using only the facts in the text. With duration at short, the episode runs 5 min 36 s.

Every date and number matches the changelog: webhook retries on 408, 425, 429 and 5xx answers with the same Idempotency-Key, in about 6 hours and with at most 30 seconds per attempt; the end of delivery on redirects, invalid HTTPS and URLs with credentials; video failures that now end at once instead of after about 90 minutes; the transcription delays of October 26 and 27; the download fix of October 28; and the intro titles of October 7. The hosts also turn the changes into advice for support, as asked.

Four sentences are wrong or go too far, and on an internal show they would mislead your own team. The changelog says up to 5 attempts in total, not "retry up to five times". The Idempotency-Key lets the receiving server spot a repeat, but it prevents nothing unless that server checks it. GET /content/Status/{id} returns the request's result, not the webhook failures. And the changelog does not say that the transcription problem is "permanently resolved".

"Use only the facts in the text" reduced the additions but did not stop them, mostly where the hosts explain why a change matters. For an internal show, ask someone who knows the product to listen before you send it, or set style to brief for a shorter episode with less room to explain. The episode took 4 min 47 s to generate and used 10 credits.

Guide for this job: Turn internal company updates into an audio podcast.

The exact request

The body that was sent to POST /content/Create, unchanged, and the same request as curl.

request.json
{
  "resources": [
    {
      "type": "text",
      "content": "AutoContent API changelog\n\n2026-11-02: Webhooks: when your endpoint answers 408, 425, 429 or a 5xx status, we send the same event again with the same `Idempotency-Key` header, up to 5 attempts in about 6 hours. Each attempt waits at most 30 seconds for your answer.\n2026-11-02: Webhooks: any other answer outside 2xx, a redirect (3xx), and a URL that is not valid HTTPS or that contains credentials end the delivery. The result of the request stays available from `GET /content/Status/{id}`.\n2026-10-29: Video: when video generation fails, the request now ends at once with a clear error message, instead of waiting about 90 minutes and then reporting a general error.\n2026-10-28: Downloads: files on data.autocontentapi.com no longer stop partway through a download.\n2026-10-27: Video: from October 26 to 27, some video requests waited up to 24 hours at the transcription step and then failed. Transcription starts at once again.\n2026-10-07: Video: long intro titles now fit inside the frame."
    }
  ],
  "outputType": "audio",
  "text": "This week's episode of our internal podcast for the whole AutoContent team: what changed for customers in these changelog entries, why each change matters, and what support should tell customers. Use only the facts in the text, and do not guess how the product works inside.",
  "duration": "short",
  "language": "English"
}
curl
curl -X POST "https://api.autocontentapi.com/content/Create" \
  -H "Authorization: Bearer $AUTOCONTENT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"resources":[{"type":"text","content":"AutoContent API changelog\n\n2026-11-02: Webhooks: when your endpoint answers 408, 425, 429 or a 5xx status, we send the same event again with the same `Idempotency-Key` header, up to 5 attempts in about 6 hours. Each attempt waits at most 30 seconds for your answer.\n2026-11-02: Webhooks: any other answer outside 2xx, a redirect (3xx), and a URL that is not valid HTTPS or that contains credentials end the delivery. The result of the request stays available from `GET /content/Status/{id}`.\n2026-10-29: Video: when video generation fails, the request now ends at once with a clear error message, instead of waiting about 90 minutes and then reporting a general error.\n2026-10-28: Downloads: files on data.autocontentapi.com no longer stop partway through a download.\n2026-10-27: Video: from October 26 to 27, some video requests waited up to 24 hours at the transcription step and then failed. Transcription starts at once again.\n2026-10-07: Video: long intro titles now fit inside the frame."}],"outputType":"audio","text":"This week'\''s episode of our internal podcast for the whole AutoContent team: what changed for customers in these changelog entries, why each change matters, and what support should tell customers. Use only the facts in the text, and do not guess how the product works inside.","duration":"short","language":"English"}'
# then poll until "status" is 100:
curl -H "Authorization: Bearer $AUTOCONTENT_API_KEY" "https://api.autocontentapi.com/content/Status/<request_id>"

How we checked it

Transcript
We transcribed the output with Whisper large-v3-turbo and compared the transcript with the source.
Language
Detected en, requested en.
Loudness
-24.3 LUFS integrated, true peak -5.6 dBTP.
Notes
Every date and number matches the changelog. Four slips, named in the teardown: "retry up to five times", the key as a promise of one processing, Status as a source of "failure results", and "permanently resolved".
Checked on
2026-11-03. Failed generations are run again and never published.