Quick Answer
The average new podcast quits after 7 episodes. The killer is not audience, it is production time - 8-12 hours per episode. This pipeline cuts it to about 3.
What you get
- Cut per-episode production from 8-12 hours to 2.5-3.5 hours
- Publish audio, transcript, and show notes together every episode instead of days apart
- Get past episode 7, where most solo podcasts quit on production time
- Ship 3-6 short clips per episode without opening a video editor from scratch
- Run the whole pipeline for $74/mo required, $130/mo with clips and AI voice
Step-by-Step Workflow
6 steps · 3 hours to set up · 3 hrs per episode ongoing
Workflow at a glance
6 steps · 3 hours setup
Research + outline
ClaudeRecord
DescriptText edit
DescriptShow notes + title
ClaudeIntro + VO
ElevenLabsDescriptPublish + clip
Descript
Opus ClipResearch + outline
Record
Text edit
Show notes + title
Intro + VO
Publish + clip
- 1
Research the episode and build the outline with Claude
Start with the topic or guest, not the microphone. Prompt Claude: 'I am recording a podcast episode on [topic] for [audience]. Build an outline: a cold-open hook, 4-6 segments each with 2-3 specific questions or talking points, and one contrarian angle most episodes on this topic miss. Keep questions open-ended, not yes/no.'
For a guest episode, add: 'Here is the guest's bio and recent work [paste]. Generate 8-10 questions that go past what they are asked in every other interview.'
The outline is your recording map. A tight outline is the biggest single lever on edit time - a rambling recording is where the hours go.
Claude - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 25 minOutput: Episode outline with hook, segments, and 8-12 questionsTools: ClaudeTip: Ask for one contrarian angle explicitly. It is what makes an episode worth listening to over the ten other shows covering the same news, and AI will not volunteer it unless you request it.
- 2
Record: locally in Descript, or in Riverside for remote guests
For a solo or in-person episode, record directly in Descript - it captures audio (and video if you want it) and transcribes as you go. For a remote guest, record in Riverside: it captures each participant on a separate local track at full quality, so a guest's bad connection does not wreck the audio.
Record to the outline but leave room to follow a good tangent. Mark mistakes by saying a keyword like 'edit that' - you will find and cut them by text later. Do not stop and re-record; keep rolling and fix it in the edit.
Descript - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 45-60 min recordingOutput: Raw recording with auto-transcript, separate tracks if remoteTools: Descript, RiversideTip: Record 3-5 seconds of room silence at the start. Descript's Studio Sound and noise removal calibrate better with a clean silence sample, and it gives you a clean handle for the intro.
- 3
Edit by deleting text, not dragging clips
Open the transcript in Descript. Edit the episode by editing the words: delete filler, cut the 'edit that' takes, remove dead air and tangents that did not land. Use Remove Filler Words to strip ums and uhs in one pass, and Studio Sound to even out audio quality.
This is where the 8-hour edit becomes a 30-45 minute read-through. Read the transcript top to bottom once, cutting as you go; resist re-listening to everything, which is what blows up the timeline.
30-45 minOutput: Edited episode audio (and video) plus a clean final transcriptTools: DescriptTip: Do one editing pass for content (what to cut) and a second quick pass for audio (filler, levels, Studio Sound). Trying to do both at once is slower and you miss things.
- 4
Generate show notes, title, and chapters from the transcript
Paste the final transcript into Claude. Prompt: 'From this podcast transcript, produce: (1) five episode-title options - specific, curiosity-driven, under 60 characters, (2) a 120-word episode description for the podcast feed, (3) chapter markers with timestamps and short labels, (4) 5 pull-quotes worth clipping, (5) a bulleted show-notes list of key points and any links or resources mentioned.'
This single prompt replaces the most-skipped, most-tedious part of podcasting. Verify the timestamps and any links against the recording before publishing.
Claude - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 15 minOutput: Titles, description, chapters, pull-quotes, and show notesTools: Claude - 5
Add the intro and optional AI voice touches
Record your intro over the clean silence handle, or generate a consistent intro/outro voiceover in ElevenLabs so every episode opens the same way without you re-recording. ElevenLabs also handles audiogram narration and, if you want reach, dubbing the episode into another language from the same transcript.
Keep the human host voice as the body of the episode; use AI voice only for the repeatable furniture (intro, outro, sponsor read) where consistency beats personality.
ElevenLabs - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 10-15 minOutput: Final episode with intro/outro in placeTools: ElevenLabs, Descript - 6
Publish to Transistor and auto-cut clips with Opus Clip
Export from Descript and upload to Transistor, which pushes the episode to Apple Podcasts, Spotify, and everywhere else via RSS. Paste in the AI-generated description and chapters. Schedule for your fixed publish day.
Then feed the episode video (or a video export) to Opus Clip. It finds the sharable moments, reframes them vertical, and adds captions - giving you 3-6 clips per episode for social without opening a video editor. Post those across the week pointing back to the full episode.
Descript - the interface you'll work in for this step. Screenshot of the tool's own UI, not our results. 20 minOutput: Published episode plus 3-6 captioned vertical clipsTools: Descript, Opus ClipTip: Publish on the same day and time every week. Podcast listening is habitual - a consistent slot compounds subscribers far more than a marginally better episode dropped whenever it is ready.
Most podcast advice obsesses over microphones and audience growth. Neither is why podcasts die. They die because a solo host spends 8-12 hours editing, writing show notes, and cutting clips for a single episode, and after seven weeks of that the calculus stops working. This pipeline attacks the time cost directly. It uses text-based editing (edit the transcript, the audio follows), AI for the research and the show-notes-and-titles busywork, and auto-clipping for social, compressing the whole loop to about three hours. The required stack is $74/mo; add remote-guest recording, AI voice, and video clips and it tops out at $130/mo.
Why solo podcast production got 4x faster
Three shifts collapsed the production timeline. First, text-based editors (Descript) let you cut filler words, ums, and whole tangents by deleting text, which turns a 2-hour timeline edit into a 30-minute read-through. Second, transcription became free and instant, so the transcript, chapters, and searchable show notes fall out of the edit rather than being separate work. Third, auto-clipping tools find the sharable moments and reframe them to vertical with captions, removing the single most tedious promotion task. The result: the parts that used to take a full day now take an afternoon, which is the difference between a podcast that survives and one that quits at episode 7. The one thing AI does not do well is the edit decisions that shape the story - what to cut, where to tighten, which tangent to keep. That judgment stays with you.
Stack cost breakdown
Public list prices as of July 2026. Optional tools are marked in the notes.
| Tool | Plan | Monthly cost | Notes |
|---|---|---|---|
| Descript | Creator | $35/mo | Required. Record, transcribe, and text-based edit. |
| Transistor | Starter | $19/mo | Required. Hosting, RSS, distribution, analytics. |
| Claude | Pro | $20/mo | Required. Research, show notes, titles, chapters. |
| Riverside | Standard | $19/mo | Optional. Remote-guest recording, separate tracks, 4K. |
| ElevenLabs | Creator | $22/mo | Optional. Intro VO, audiogram narration, dubbing. |
| Opus Clip | Starter | $15/mo | Optional. Auto-cut vertical video clips for social. |
| Total | $74 - $130/mo($74 required, $130 with optional tools) | ||
Email me this stack as a checklist
Every tool, the plan to pick, and the monthly cost - in your inbox.
Real usage
What people actually run
No usage reports yet - be the first to share what you run. Tell us your real stack, your actual monthly cost, and any tools you swapped.
The prompts that do the busywork
The two prompts below are the load-bearing ones. The first shapes the recording so the edit is short; the second turns the transcript into everything you publish alongside the audio.
Guest-episode question prompt (Claude)
I am interviewing [guest name], [one-line bio], on [topic]. Here is their recent work: [paste links or summary]. Generate 8-10 open-ended questions that: (1) go past what they are asked in every other interview, (2) build on each other in a logical arc, (3) include one question that gently challenges a position they are known for. No yes/no questions. Add a cold-open hook I can read before the first question.
Note: The 'gently challenges' line is what produces a memorable moment. Interviews where the host only lobs softballs are the ones nobody clips or shares.
Transcript-to-publish prompt (Claude)
From this podcast transcript, produce: (1) five title options, specific and under 60 characters; (2) a 120-word feed description; (3) chapter markers with timestamps and short labels; (4) five pull-quotes worth turning into clips, with their approximate timestamps; (5) show notes as bullets, including any links, tools, or people mentioned. Transcript: [paste].
Note: Always verify timestamps and mentioned links against the recording - the transcript timing is close but not frame-accurate, and a wrong link in show notes reads as sloppy.
Adjust for Your Situation
If you are audio-only on a tight budget
Drop Descript to Adobe Podcast (free enhance and transcription) plus Audacity for cutting, and drop Opus Clip (which needs video). Keep Transistor and Claude. Required cost falls to about $39/mo (Transistor plus Claude). You trade text-based editing convenience for a slower timeline edit.
If most episodes are remote interviews
Make Riverside required, not optional - separate local tracks per guest are the difference between usable audio and a re-record when someone's wifi drops. Record in Riverside, then bring the tracks into Descript for the text-based edit. This raises your required stack to about $93/mo.
If you want a video podcast (YouTube first)
Record video in Riverside or Descript, edit in Descript, publish the full video to YouTube and the audio to Transistor. Lean harder on Opus Clip for clips (Pro at $29/mo for more processing minutes). YouTube becomes your discovery engine and the audio feed becomes the subscriber base.
Swap options
Drop-in substitutions if a tool does not fit your budget or stack. These trade cost or effort for the recommended setup.
| Swap out | Use instead | When |
|---|---|---|
| Descript | Adobe Podcast (free) plus Audacity | You are audio-only, on a tight budget, and willing to edit on a timeline instead of a transcript |
| Transistor | Buzzsprout or Captivate | You want per-episode pricing (Buzzsprout) or creator-network tooling (Captivate) |
| Opus Clip | Castmagic | You want clips plus AI show notes and social copy from one upload and will pay $39/mo for it |
| Riverside | Zoom plus local track recording | You record remote guests occasionally and cannot justify a second recording tool |
Common Pitfalls
- Recording without a tight outline. A rambling take is where the editing hours disappear. Spend 25 minutes on the outline to save two hours on the edit.
- Re-listening to the whole episode while editing. Read the transcript instead. Re-listening is the single biggest time sink in solo production.
- Publishing audio now and show notes 'later.' Later never comes. Generate notes, chapters, and title in the same session from the transcript.
- Using AI voice for the whole episode. Listeners subscribe to a person. Use ElevenLabs only for the repeatable furniture (intro, outro, sponsor read).
- Skipping separate tracks for remote guests. One bad connection on a single mixed track means a re-record. Riverside's per-guest local tracks are cheap insurance.
- Inconsistent publish days. Podcast listening is a habit. Moving the drop day resets the habit and flattens your growth.
When to change the format instead of quitting
If you hit episode 10-12 and downloads are flat under ~50 per episode with no upward trend, do not just push harder on the same format - the problem is usually the concept, not the effort. Before quitting, try one structural change: switch from solo monologue to guest interviews (or the reverse), tighten episodes to under 25 minutes, or narrow the topic to something more specific than your current scope. Give the new format four episodes. If downloads still do not move, the show concept is not landing with a findable audience, and no amount of production polish fixes a positioning problem. Abandon the concept, keep the pipeline, and relaunch with a sharper premise - the setup you built transfers directly.
Frequently Asked Questions
Do I really need both Descript and a separate host like Transistor?
Can I do this entirely free?
How much does AI actually save versus doing it manually?
Will AI-generated show notes hurt discoverability?
Is Riverside worth it if I record mostly solo?
How do the clips actually help the podcast grow?
How we built this workflow
Prices verified July 2026 from each vendor's pricing page: Descript Creator $35/mo, Transistor Starter $19/mo, Claude Pro $20/mo, Riverside Standard $19/mo, ElevenLabs Creator $22/mo, Opus Clip Starter $15/mo. The 8-12 hour manual baseline and 2.5-3.5 hour AI-assisted range reflect typical solo production; interview shows with heavy editing run longer, tight solo monologues run shorter.
Last updated July 7, 2026; prices verified at publication.