Podcasting · checked 2026-10-19

Can I vibecode Castmagic?

YES· one-shottable

Castmagic is a transcript plus a set of prompts. A local app runs Whisper on an audio file, then runs a folder of editable prompt templates over the transcript for show notes, chapters, key quotes, a newsletter draft and social posts, and saves the results next to the recording. That is one sitting with faster-whisper and any language model. What you give up is the studio around it: clip and audiogram rendering with captions, shared team spaces, one-click imports from YouTube, Spotify and Zoom, and an API. The quality of the outputs then depends on prompts you tune yourself.

$29/moprice

$348you'd save a year

one sittingbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyIntermediate · Some system setup or keys.

TimeOne to two hours for the first version

Claude Code, Codex or CursorAny of them can run the prompt.
Python 3.11 and ffmpegfaster-whisper downloads its model on first run; the small model is a few hundred megabytes.
A language model endpointA hosted API key, or Ollama running on your own machine.
Your Castmagic transcripts and outputsExport them first so your best prompts and results carry over.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a local audio-to-content tool to replace Castmagic for my podcast episodes. Requirements:

- Python 3.11 FastAPI backend with a server-rendered UI (Jinja2 + htmx). `./run.sh` serves
http://127.0.0.1:4720. ffmpeg must be on the PATH; exit with a clear message if not.
- Import: drop in an mp3, m4a, wav, mp4 or mov. Copy it to ./data/episodes/<id>/ and extract 16 kHz
mono audio with ffmpeg. No URL imports.
- Transcription: faster-whisper with word timestamps and the VAD filter (model from .env, default
`small`) and a progress bar. Save transcript.json plus .srt and .txt, and show the transcript
next to an audio player: click a word to seek, edit a word to update the files.
- Prompts: a ./prompts/ folder of Markdown templates with a front-matter title, starting with
show-notes.md, chapters.md, key-quotes.md, newsletter.md, linkedin-post.md and x-thread.md.
The UI lists, edits and adds them. Placeholders: {{transcript}}, {{title}}, {{guest}}.
- Generate: run the selected prompts over the transcript against an OpenAI-compatible endpoint
(LLM_BASE_URL, LLM_MODEL, LLM_API_KEY in .env; local Ollama works). Split long transcripts into
chunks of about 6,000 tokens and merge the results. Save each output as
./data/episodes/<id>/outputs/<prompt>.md, keep earlier versions, and print the token count of every run.
- Chat: a box that answers questions about one episode and quotes the timestamp of each passage
it relies on.
- Chapter check: every chapter timestamp must be inside the audio length and in ascending order;
drop or fix the bad ones and say so.
- Export: a ZIP per episode with the transcript (.txt, .srt, .vtt), all outputs and the audio.
- Single user. No accounts, no telemetry, local-first; the only outbound call is the language model
endpoint, and with Ollama the app works offline.
- Out of scope: audiograms and clip rendering, URL and RSS imports, teams, an API.
- README: the models to download, the .env keys, how to write a prompt template.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Castmagic.

Add feed import to the audio-to-content tool in this folder. Requirements:

- FEED_URL in .env; a command `import-feed` that reads the RSS feed with `feedparser`, lists
  episodes not yet imported and downloads the enclosures into ./data/episodes/.
- Store the episode title, date and description as metadata and pass them into the prompt
  placeholders {{title}} and {{date}}.
- Skip enclosures larger than 500 MB and say so.
- Update the README with a cron example.

Add clip rendering to the audio-to-content tool in this folder. Requirements:

- Select a start and end in the transcript view and render an MP4 with ffmpeg: a 1080x1080
  background in a color from .env, an animated waveform (showwaves) and the spoken words as
  burned-in ASS captions timed from the word timestamps.
- Cap clips at 90 seconds and show the output path.
- Provide the cover image path as an optional setting.
- Update the README.

Add a style guide to the audio-to-content tool in this folder. Requirements:

- A ./style/ folder where I drop up to ten past posts as Markdown files.
- A "Build style guide" button that asks the language model for a one-page description of my
  voice (sentence length, openings, words I avoid) from those files and saves it as
  ./style/guide.md, which I can edit.
- Prepend the guide to every prompt run and show it in the output header so I can see it was used.
- Update the README.

What you lose

Audiogram and clip rendering from the Studio
Imports from YouTube, TikTok, Instagram, Spotify, Zoom and Google Drive
Team spaces, shared recordings and an onboarding call
A community prompt library, an API and webhooks
AI image generation and a media library for the clips and covers

Why people still pay

Podcasters and agencies pay to skip prompt tuning: upload an episode, get show notes, chapters, quotes and social drafts in a consistent format, plus clips to post. The subscription buys time saved per episode, and because hours are a library total, heavier users climb the tiers.

Moatexecution polish integrations

Do the math

Castmagic or your own copy?

Over

Your usage

Count my time3h ×$/h

Castmagic, 1 yr$348

Your own copy, 1 yr$53running + build session

You keep$295pays for itself in the first month

Estimate · Transcription runs on your machine. The only running cost is language model tokens, typically cents per episode on a small hosted model and nothing with Ollama. The build is token usage on your agent plan.

Leaving Castmagic

Take your data, then cancel.

1 · Take your data out of Castmagic

1Transcripts: the API docs read transcripts as json, txt, vtt, srt, csv, pdf or docx. The docs don't name the in-app menu, so look for the export option on each recording.
2Generated outputs (show notes, posts, quotes): the API lists a recording's AI-generated content in markdown through its outputs endpoint, which needs a plan with API access (Starter and up). We found no documented bulk download for Hobby, so copy the outputs you want per recording.
3Original media is yours: the terms say users keep ownership of what they upload, but keep your own copies of the audio and video.

2 · Cancel Castmagic

1Cancel from your account settings or by contacting support; the terms allow both. We found no help article with the exact click path.
2The subscription stays active until the end of the current billing period.
3Payments are non-refundable and partial months aren't refunded.

The terms don't say what happens to your library after you cancel, so export transcripts and outputs first.
Annual plans bill once a year (Hobby is $239 up front), and the pricing page advertises no free trial, so start monthly if you are testing.
Plans can be upgraded or downgraded at any time per the pricing page; downgrading is the reversible option compared with leaving.
Keep your best custom prompts: they are text you wrote, so copy them out to reuse in a local tool.

Checked 2026-10-19 against Castmagic's docs: 1, 2, 3, 4

The escape hatch

Don't feel like building it? These are already free.

ScriberrOpen sourceMIT self-hosted, offline audio transcription with local Whisper, Parakeet or Canary models; the README notes the maintainer paused and resumed work.★ 3.1k · active Sep 2026 · docker to self-host
Web, serverVisit ↗BuzzOpen sourceMIT desktop app that transcribes and translates audio offline with Whisper; transcripts only, no show-notes prompts.★ 22k · active Oct 2026 · one-click install
macOS, Windows, LinuxVisit ↗whisper.cppOpen sourceMIT C/C++ port of Whisper that runs on laptops; the transcription engine only, with no content generation.★ 54k · active Oct 2026 · self-host, real ops
CLIVisit ↗

All 4 free alternatives to Castmagic ›

The numbers

Castmagic pricing

Plan	Monthly	Annual, per month	What you get
Hobby	$29	$19	30 transcribed hours (library total), up to 3 spaces, 50 GB storage, 3 seats, standard support, no API.
Starter	$69	$48	100 transcribed hours, up to 10 spaces, 250 GB, 5 seats, API access and an onboarding call.
Team	$199	$139	400 transcribed hours, up to 25 spaces, 1 TB, unlimited seats with 5 included, priority support, API and an onboarding call.
Business & Scale	—	$699	Custom volume, seats and storage, SLA and dedicated support for studios, platforms and publishers.

Free tier: The pricing page mentions no free plan or trial; the buttons lead to sign-up. Transcribed hours are a library total, not a monthly allowance.
Billing: Monthly or annual; annual is advertised as about 30% cheaper and is the page default. Plans can be upgraded or downgraded at any time and extra hours bought. Under the terms, payments are non-refundable and partial months aren't refunded; cancelling keeps the subscription active to the end of the paid period. Prices use dollar signs without a currency code.
Hidden costs: Hours count the whole library, so a long back catalogue uses the allowance even if you upload nothing new. The terms say overage fees may apply, and extra hours can be bought. API access starts at Starter, and Team seats beyond five cost extra.
Source: www.castmagic.io

Also on the list

Descript$24/mo · Audio & videoKINDA

Riverside$29/mo · PodcastingKINDA

Granola$14/mo · Meeting notesYES

Questions

Castmagic, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

Yes. An AI coding agent such as Claude Code, Codex or Cursor can build a usable personal version in one session with the prompt on this page, running on your own machine.

About $29 a month on the Hobby plan, $348 a year (checked 2026-10-19).

Audiogram and clip rendering from the Studio; Imports from YouTube, TikTok, Instagram, Spotify, Zoom and Google Drive; Team spaces, shared recordings and an onboarding call; A community prompt library, an API and webhooks; AI image generation and a media library for the clips and covers. If any of those matter to you, keep paying.