Podcasting

A podcast is the cheapest way to be in someone’s ears every week, and the people who listen to yours are the people who buy your courses. On Omumu the podcast lives on your own site: one show with a public feed that Apple Podcasts, Spotify and every other app read, episodes you upload as plain audio files and the platform masters, transcribes and serves, and an editor that lets you cut an “um” by selecting the word.

This chapter is the show from nothing to listed. Your assistant can publish episodes for you; the show itself, and the editing, are admin work, and the chapter says which is which. A path such as Products → Podcast names a group in the admin’s navigation and the page under it.

Concepts

One show, one feed. A site has one podcast show. The show carries what directories need — title, description, language, a category from Apple’s list, author and owner, an explicit flag, a show type — and a slug that becomes the public address: the show page, each episode’s page, and the feed at /podcast/<slug>/feed.rss. You submit that feed address once per directory; after that, new episodes reach every app on their own.

The cover. Directories require square cover art, at least 1400 and at most 3000 pixels on a side. You can pick an uploaded image, or let the show page draft covers for you: a template, a short prompt for the background, and the show’s title and author composited on top at 3000 × 3000. Publishing an episode while the show has no cover warns you, because the feed would be refused by the directories.

An episode is a file you upload. MP3, M4A (an iPhone voice memo) or WAV. On upload the platform masters it — brings it to podcast loudness — and the episode sits in processing until that is done, then ready. Omumu’s own chain does the mastering unless the site has an Auphonic connection: then your own Auphonic account masters every new upload (adaptive leveling, automatic EQ, bandwidth extension) on your own Auphonic credits, and the episode page says which of the two mastered it. If Auphonic cannot master a new upload — credits used up, key revoked, Auphonic unreachable — Omumu’s chain masters it instead, the episode says why, and its page warns before you publish. Noise reduction is a per-episode choice, off by default, for a recording made outdoors or in a noisy room. Publishing stamps the publication date, locks the audio file (subscribers’ apps may already have fetched it) and puts the episode in the feed. You don’t have to wait for mastering to publish: publishing an episode that is still processing records that it should go live, and it publishes itself the moment mastering ends, dated then — also when Omumu’s chain had to master it instead of Auphonic, and then you get an email saying why, with a link to the episode. A failed mastering publishes nothing, keeps that request for the retry and emails you the reason with a link to retry it, and until the episode is live you can cancel it. A publish that goes as planned sends no email. Mastering can also wait: Upload without mastering leaves the episode unmastered, so no mastering is spent on a recording you will trim first. It is mastered when you click Master, when you publish it (mastered first, then live), or when you open it in the editor and accept mastering and transcription together. Mastering is deferred, never skipped: nothing reaches the feed unmastered. A failed mastering can be retried, and any episode can be re-mastered: the original upload is mastered again with the current chain, and a published episode keeps playing its current audio until the new master is ready. Episodes have a type (full, trailer, bonus), a season and number, an explicit flag that can inherit the show’s, and optional artwork of their own.

Transcripts come with the episode. Every episode is transcribed by speech-to-text — at publish, or earlier if you open the editor and accept the cost in plan minutes. The transcript ships in the feed in the formats podcast apps read and is shown on the episode’s public page, with speaker labels. One transcription serves every round of editing; a re-master of a published episode transcribes it again, on plan minutes, because the transcript is aligned to the audio it was made from.

Editing is editing the words. The episode editor plays the audio with the transcript highlighted word by word, under a waveform of the whole recording, the audio no word covers included, with the playback position and every cut shown in both. Select a passage and cut it, or drag across the waveform to cut audio no word covers, such as music before the first word or a jingle after the last (a cut that reaches into a word keeps or removes the whole word, so the transcript always matches the audio); tighten long pauses, one or all at once; correct a misheard word (text only; the audio is untouched). Edits are saved as you go but change nothing until you apply, which renders the new audio and makes it the version listeners get. The original recording is kept, so a cut can be undone in a later session and renders never stack. Editing is a plan feature of its own.

One recording, several episodes. A long conversation can be split into several episodes at points you choose. Each part is a draft of its own with its slice of the master and of the transcript, so nothing is mastered or transcribed again, and the cuts you already made carry into the parts. The original stays as a locked source that links to its parts, so the same audio never reaches the feed twice. Only an unpublished episode that is mastered and transcribed can be split, and the split can be undone until a part is published.

Hands-on

Create the show

Products → Podcast starts with the show. Mira fills in the title, a description, the language, a category (Education → How To), her name as author and owner, and a slug. Then the cover: she types a background prompt, picks a template, generates a draft, and accepts the one she likes — it becomes a 3000 × 3000 cover with the title and author set on it. The page shows the feed address and links to the distribution checklist and analytics.

There are no agent tools for creating or editing the show; omumu_podcast_show_get reads it — title, slug, description, language and feed address — once it exists.

Publish an episode

Mira records a 20-minute episode on her phone and says: “Publish this as episode 2 of my podcast, ‘Automate one task’, with these show notes.”

The assistant uploads the file straight to storage with omumu_resource_upload_presign, creates the episode with omumu_podcast_episode_create (the audio resource, title, show notes, season and number, type, explicit flag) and publishes it at once with omumu_podcast_episode_publish. The episode is still processing, so the call records that it should go live, and omumu_podcast_episode_get reports publishIntent until mastering ends and the episode publishes itself; omumu_podcast_episode_publish_cancel takes it back until then. omumu_podcast_episode_list shows every episode and its status; omumu_podcast_episode_retry re-runs a failed mastering. For a recording she will trim first, deferMastering on omumu_podcast_episode_create leaves the episode unmastered; omumu_podcast_episode_master starts its mastering when she is ready, and publishing it masters it first.

In the admin: the show page’s New episode takes the audio file, title, show notes, type, season and number, rating and artwork, and Upload without mastering holds the mastering back; Create episode lands you back on the show with the episode in its list, where an unmastered one has Master beside it. The episode’s own page shows its status, lets you edit the details, generate episode artwork (the show’s background with this episode’s title, no AI involved), see the transcript’s state, and Publish in a dialog that reminds you the date is stamped and the audio locked. An episode your assistant asked to publish while it was still being mastered says so — This episode will be published as soon as it is mastered — with Cancel publishing beside it. An unmastered episode’s page offers Master and Publish, which masters it first and then publishes it.

Clean up the audio

Mira listens back and hears three “um”s and a long pause. On the episode page she opens the Episode editor. If the episode is not transcribed yet, the editor says how many plan minutes it will cost and asks before it starts. If it is not mastered yet either, the editor offers Master and transcribe, with both costs first: Auphonic credits for each minute of the recording when Auphonic masters for the site (included otherwise), and a plan minute for each minute of the transcript. Accepting masters the episode, then transcribes it, and the editor opens when the transcript is ready. The first time she opens it, the waveform above the transcript says it is being drawn, and shows the whole recording a moment later. Then she clicks a word to listen from there, or clicks the waveform to jump to that moment (the arrow keys do the same on it), and Zoom in until a tenth of a second is easy to see. She shift-clicks the first and last word of a passage and Cuts it, uses Tighten all pauses with a threshold in seconds, and each cut and tightened pause shows hatched on the waveform at once. Her Auphonic free plan put a jingle after her last word, so she drags across the waveform from the end of that word past the end of the recording: the jingle is hatched, listed under the waveform as an audio cut with its own Remove, and the duration drops to where she stopped talking. A drag that ends inside a word lands on that word’s nearer edge and strikes it in the transcript, or keeps it whole. On a phone, or without a mouse, Start an audio cut here at one moment and End the audio cut here at another does the same. She double-clicks a misheard word to correct it, and finally Apply edits. The feed and the episode page switch to the rendered audio and the corrected transcript; the original stays behind the scenes for the next round.

Editing itself has no agent tools.

Mira’s early episodes were recorded in a café and sound thin next to the new ones. Under Integrations → Auphonic she pastes the API key from her Auphonic account; Omumu checks it on save, and from then on every new upload is mastered by Auphonic on her credits. The page also says that Auphonic’s free plan adds a short jingle to the start and end of every episode it masters, and that paid Auphonic credits remove it. Omumu cannot see which plan she is on, so everyone gets the same notice, and it never removes the jingle itself: an episode published before mastering ends goes live with it, and one still unpublished she can cut on the waveform, as above. For the old episodes she opens one and clicks Re-master: the original upload is mastered again with the current sound, listeners hear the current audio until the new master lands, and then the feed switches over and a fresh transcript replaces the old one. For the whole back catalogue the show page’s Re-master all lists how many episodes it covers, which ones it leaves out and why (edits applied against the current master’s word timings, a re-master already running, already mastered with the current sound, not mastered yet), how many Auphonic minutes it will use, and how many transcription minutes the fresh transcripts will take from the plan’s pool. Re-masters run one at a time. The café episodes also get Remove background noise switched on, which is an episode’s own setting, chosen at upload or when re-mastering; it helps a noisy recording and can make a clean one sound harsh.

Her assistant can do the re-mastering: “Re-master episode 2 with noise reduction” is omumu_podcast_episode_remaster, and omumu_podcast_episode_get and _list report the re-master’s state, what mastered the current audio, why Auphonic did not, and whether noise reduction is on. Connecting Auphonic is admin-only, never over MCP.

Split a long conversation

Mira recorded a 50-minute conversation with two guests, and it is really three episodes. In the editor, under the waveform, Split into episodes takes the points to split at. She clicks the word where the second topic starts and presses Add the playback position, does the same where the third starts, and the field reads 16:42.3, 33:10.8; she could type the times as well. A point that lands inside a word moves to the word’s nearer edge, so no word is cut in half. Split cuts the parts while the editor follows along, then lists them: (1/3), (2/3) and (3/3) after the episode’s title, each a draft with its slice of the master and of the transcript, the show notes and the artwork. Nothing is mastered or transcribed again, so the split uses no Auphonic credits and no plan minutes. The cuts she had already made carry into the parts, a cut across a split point becoming one cut in each, and she edits, applies and publishes each part on its own, days apart. The original stays as a locked source: it cannot be published, edited or re-mastered, the episode list leaves it out, and its page links to its parts. A part cannot be re-mastered either, because its timings belong to the original’s master.

If a cut point turns out wrong, the original’s page offers Undo split until one of the parts is published. Undoing deletes the parts, and any edits made in them are lost, which the confirmation says before anything happens; the original is a draft again with the cuts it had before, ready to be split at other points. Undoing costs nothing either. It is not available while a part is being mastered or having its edits applied, or once a part is published or has been split in turn; the page then says why the split cannot be undone.

Splitting has no agent tools.

Get listed

Products → Podcast → Distribution checklist lists the directories in the order to submit to — Apple Podcasts, Spotify, Podcast Index, Pocket Casts, YouTube Music, Amazon Music — each with a link to its submission page, a note on how long it takes, and a status you set yourself (not submitted, submitted, live). Apple may ask you to verify ownership; paste the token it shows into the checklist and it appears in your feed. Spotify mails its verification code to the show’s owner address.

See who listens

Products → Podcast → Analytics counts downloads (audio requests from apps and browsers, de-duplicated per listener per day) and plays in the web player, per episode and for a date range. The counts are estimates aligned with the industry’s measurement rules, not certified figures.

What you cannot do yet

Stated plainly: one show per site; episodes publish now or as soon as they are mastered, never at a time you choose; an episode’s audio cannot be replaced after publishing; your assistant cannot create or change the show, edit an episode’s details, unpublish or delete one, or run the editor; once an audio cut is made its edges cannot be dragged, only the whole cut removed; a split cannot be undone once a part is published, and a published episode cannot be split; and the episode’s admin page has no player — listen in the editor or on the public page.

Reference

omumu_podcast_episode_create

Creates a draft episode on the site’s show from an already-uploaded audio resource (upload the file first with omumu_resource_upload_presign; mp3, m4a, and wav are accepted). Mastering starts automatically and the episode is returned in PROCESSING status; publish it right away and it goes live when mastering ends, or poll omumu_podcast_episode_get until it reaches READY, then publish. With deferMastering the episode is returned UNMASTERED and nothing is mastered yet: start mastering with omumu_podcast_episode_master, or publish it and it is mastered first, then goes live.

• Write category: standard write
• Access: requires the PODCAST_WRITE permission

omumu_podcast_episode_get

Gets one episode by id with its processing status (UNMASTERED, PROCESSING, READY, FAILED, PUBLISHED, SPLITTING while being split into other episodes, SPLIT once it has been), duration, error message when failed, pubDate when published, the latest re-master’s status (NONE, PENDING, FAILED) with its error when failed, masteringMethod: what made the current master (auphonic, omumu_chain, or null when not mastered yet or mastered by an older Omumu chain), masteringFallback: why Auphonic could not master it when Omumu’s chain made it instead (out_of_credits, key_rejected, production_failed, unreachable, timed_out), null when it did not fall back, noiseReduction: whether the episode is mastered with noise reduction, and publishIntent: whether it publishes itself when mastering ends.

• Write category: read-only
• Access: requires the PODCAST_READ permission

omumu_podcast_episode_list

Lists the show’s episodes with their processing status (UNMASTERED, PROCESSING, READY, FAILED, PUBLISHED, SPLITTING while being split into other episodes, SPLIT once it has been), duration, error message when failed, pubDate when published, the latest re-master’s status (NONE, PENDING, FAILED) with its error when failed, masteringMethod: what made the current master (auphonic for the site’s connected Auphonic account, omumu_chain for Omumu’s own chain, null when not mastered yet or mastered by an older Omumu chain), masteringFallback: why Auphonic could not master it when Omumu’s chain made it instead (out_of_credits, key_rejected, production_failed, unreachable, timed_out), null when it did not fall back, noiseReduction: whether the episode is mastered with noise reduction, and publishIntent: whether it publishes itself when mastering ends. The source of a split is listed too; it cannot be edited, published or re-mastered, and its parts are the episodes now.

• Write category: read-only
• Access: requires the PODCAST_READ permission

No parameters.

omumu_podcast_episode_master

Starts mastering an UNMASTERED episode (one created with deferMastering), returning it in PROCESSING status; it reaches READY when mastering ends. Episodes in any other status are refused with their current status: their mastering has already started.

• Write category: standard write
• Access: requires the PODCAST_WRITE permission

omumu_podcast_episode_publish

Publishes a READY episode: the pubDate is stamped now and the episode appears in the RSS feed. A PROCESSING episode records a publish intent instead and is returned unpublished with publishIntent true: it publishes itself when mastering ends, with the pubDate stamped then, and the intent survives a failed mastering and its retry. An UNMASTERED episode starts mastering with a publish intent the same way, so it goes live mastered. Cancel it with omumu_podcast_episode_publish_cancel until then. A FAILED episode is refused with its status.

• Write category: standard write
• Access: requires the PODCAST_WRITE permission

omumu_podcast_episode_publish_cancel

Cancels the publish intent of an episode that is still being mastered, or whose mastering failed: it stays a draft when mastering ends. Refused once the episode is published.

• Write category: standard write
• Access: requires the PODCAST_WRITE permission

omumu_podcast_episode_remaster

Re-masters an episode’s original upload with the current mastering chain. A READY episode returns to PROCESSING. A PUBLISHED episode stays in the feed with its current audio while remasterStatus is PENDING; when the new master lands it replaces the audio and a fresh transcription run starts, and a failure sets remasterStatus FAILED with remasterError and leaves the published audio unchanged. Refused, with the reason, for episodes with applied transcript edits, UNMASTERED, PROCESSING or FAILED episodes, episodes with a re-master already pending, the source of a split (SPLITTING or SPLIT), and the parts of a split.

• Write category: standard write
• Access: requires the PODCAST_WRITE permission

omumu_podcast_episode_retry

Re-runs mastering on a FAILED episode, returning it to PROCESSING. Episodes in any other status are refused with their current status.

• Write category: standard write
• Access: requires the PODCAST_WRITE permission

omumu_podcast_show_get

Gets the site’s podcast show: title, slug, description, language, and the public RSS feed URL. Errors if the site has no show yet (shows are created in the podcast admin).

• Write category: read-only
• Access: requires the PODCAST_READ permission

No parameters.