From this week, every essay on the Dock blog ships as both a long-form article AND a 10-to-15-minute audio episode. Same writer, same content, two surfaces. If you'd rather read, the page is there. If you'd rather listen — driving, walking, ironing a shirt — there's a play button on every post and a feed at trydock.ai/blog/feed.xml you can drop into any podcast app.
This is the kind of feature that sounds tiny in a launch post and is meaningful in actual use. The reason we built it is straightforward: most of our readers tell us they don't have time to read a 4,000-word essay at their desk. They have time on the commute, during a walk, between meetings. So the essay has to meet them on those surfaces too, not just the ones with a screen.
How it sounds
The voice is Joseff Novak via ElevenLabs — warm, conversational, podcast-host register, not the synthetic flat-affect TTS you remember from a few years ago. We considered a celebrity voice and a custom-trained voice; both felt like overkill. Joseff is good enough that listeners stop noticing it's TTS within about 30 seconds.
The structure of every episode is the same four phases:
- Sizzle reel — three to five hook lines from the essay, played over an energetic music bed. The cold-open style that podcasts use to tell you what you're about to hear.
- Show intro — a category-specific stinger: "You're listening to Dock Architecture" / "Dock Essays" / "Dock Field Notes" depending on the post's type. Same pattern as every podcast with multiple show formats.
- The read — the full essay narrated under a quiet ambient bed. Joseff's pace is about 200 words per minute, so a 3,500-word pillar runs about 17 minutes.
- Outro — a short subscribe ask. "We ship new tracks on AI every week. Subscribe now."
Each phase is mixed independently and concatenated with brief silences. The music ducks under voice via sidechain compression so you never have to fight to hear the words. The mastering chain (compand + presence EQ at 3 kHz + sample-rate lock) gives the voice the broadcast quality that long-form audio expects.
The technical pipeline
The render is fully automated. For every essay we publish:
- Opus rewrites the article body into a podcast script — flowing prose instead of headings and lists, smooths em-dashes and abbreviations TTS reads weird, replaces code blocks with one-sentence paraphrases. Same content, voiced for the ear.
- ElevenLabs TTS turns the script into audio with the Joseff Novak voice. Per-post hooks for the sizzle get separate voice settings tuned for broadcast energy; the main read uses calmer narrator settings.
- ffmpeg mixes the four phases, ducks the music, applies the mastering chain.
- The final MP3 lands on Vercel Blob, the manifest at src/lib/blog-audio.ts updates with the URL, duration, and byte size.
Total cost is around $1.50 per essay (Opus + ElevenLabs). Total wall time is about three minutes from "essay merged" to "audio live." We re-render automatically when the essay's text changes meaningfully.
How to subscribe
Three options, pick whichever fits how you already listen:
- Add the feed directly: paste https://trydock.ai/blog/feed.xml into your podcast app's "Add by URL" / "Add by RSS" option. Works in Apple Podcasts, Pocket Casts, Overcast, Castro, AntennaPod, any standard reader.
- Search the directories: we'll be in Apple Podcasts and Spotify within the next two weeks. Search "Dock" once they index us.
- Just hit play on a post: every essay page has a play button at the top. The audio streams from the same Blob URL the feed points at, no app required.
What's actually in the feed today
The 37 essays we have so far, each with a 10-to-15-minute audio version. The pillars (5 of them, each in the 12-15 minute range) and the spoke essays underneath (most 5-10 minutes). New essays will appear in the feed within minutes of publication.
What's next
The audio companion is one of the things we're going to keep investing in instead of treating as a one-off launch. Specifically:
- Customer episodes — once we have customers willing to record, conversations about how they're running agents will become the second show format, alongside the essay narrations.
- Subscriber-only episodes — for the Pro tier, we'll do quarterly deep-dives that don't ship as posts. Audio-first content that justifies the subscription bump.
- Better voice work — Joseff is great; we're also experimenting with multi-voice production (the pillar essays read by different voices depending on what they're arguing).
- Transcripts on the audio page — for accessibility and for the SEO win. Audio transcripts are some of the best long-tail keyword density we'll ever generate.
The audio companions have been quietly running for weeks. The reason this post exists is to make them findable — until now there was no announcement, no podcast-format feed, no obvious subscribe path. From today, all three exist.
If you've been reading the essays, the audio is the same content, just made for a different time of day. Try one on your next walk and let us know what we should fix.