Turn a 60-minute podcast into 30 viral clips
Drop a podcast episode (YouTube URL, mp3, or video file). Whisper transcribes locally, AI ranks 5–30 candidate clips by virality score with hook patterns already classified, and each pick exports as a vertical 9:16 short with word-by-word captions baked in. Free Opus Clip alternative built for podcasters.
The podcaster workflow
- Drop the episode. Paste a YouTube podcast URL or import an mp3 / mp4 directly. yt-dlp pulls the source video at the highest available quality.
- Whisper transcribes locally. A 60-minute episode transcribes in 3–6 minutes on a mid-range laptop CPU (faster-whisper int8). Your audio never leaves your machine.
- AI ranks the clips. Every candidate segment gets a virality score — engagement signals, hook strength, narrative arc, emotional valence — and the hook pattern is classified per clip (curiosity gap / contrarian claim / story open / data drop / etc.).
- Pick from the ranked list. Open Clip Studio, scroll the 5–30 ranked candidates with their scores + hook tags + preview transcripts. Click to keep the ones you want.
- Export vertical with captions. Each kept clip exports as a 9:16 mp4 with face-tracking (speaker centered), word-by-word ASS captions in your preferred style, and platform metadata (title + description + tags) pre-drafted.
Why this works for podcasters specifically
- Local Whisper unblocks long-form content. Cloud transcription services meter minutes — a 90-minute episode hits cap on most plans. ViralMint runs Whisper locally with no per-minute cost; transcribe 5-hour episodes if you want.
- AI virality scoring beats moment detection. Naive clippers surface "loud moments" or laughter peaks. ViralMint's scoring weighs hook patterns (the first 10 seconds), emotional arc, and the kind of contrarian takes that travel on TikTok / Shorts. The hook-pattern classifier tells you why a clip should rank.
- Face-tracking reframe for video podcasts. If you're shooting 16:9 (most video podcasts are), the auto-reframe to 9:16 uses MediaPipe face-tracking to keep whoever's currently speaking centered in the vertical crop, not the static center of the original frame.
- Word-by-word captions on every export. The TikTok-viral caption style (one or two words at a time, synced to the speaker) is the format that increases watch time on Shorts. Sentence-level subtitles read as dated by 2026 standards.
What it costs to clip 10 episodes
Realistic example: a weekly podcast at 60 minutes per episode, extracted across a typical month (4 episodes), keeping ~6 clips per episode for posting.
- Download (yt-dlp): Free, runs locally.
- Transcription (Whisper): Free, runs locally. ~4 minutes wall-time per 60-min episode.
- AI virality scoring: ~$0.02 per candidate clip × 15 candidates × 4 episodes = ~$1.20/month.
- Caption rendering (FFmpeg): Free, runs locally.
- Reframe + export: Free, runs locally.
Total: under $2/month for a weekly-podcast workflow. Compare to Opus Clip's $19/mo starter plan or Headliner's $13/mo audiogram plan — both monthly subscriptions regardless of usage.
What it doesn't do
- No podcast-host auto-publish. ViralMint generates the mp4 + metadata; you copy-paste and post manually. If you need one-click publishing from Megaphone / Libsyn directly to TikTok, Headliner's integrations win.
- No animated waveform overlay. If you specifically want the audiogram aesthetic (waveform animation under episode cover art), Headliner is purpose-built for that. ViralMint's output is video-first.
- No real-time collaboration. Single-user desktop app. Multi-user review workflows aren't a feature.
Frequently asked
How does ViralMint pick which podcast moments to clip?
After Whisper transcribes the episode locally, an AI scoring pass ranks every candidate segment by hook strength, narrative arc, emotional valence, and outlier signal. The top 5–30 clips come back ranked, with each clip's hook type already classified (curiosity gap / contrarian claim / story open / etc.) so you pick from a ranked list instead of scrubbing the timeline.
What podcast sources work with ViralMint?
Anything yt-dlp can download (YouTube podcast channels, Spotify-via-RSS where the audio is hosted on a fetchable URL, direct mp3/mp4 imports) plus any audio file you drop into the desktop app's import flow. There's no podcast-host integration today (no Megaphone/Libsyn API hooks), so the flow is: paste the YouTube/RSS URL or drop the mp3, ViralMint handles the rest locally.
How much does it cost per podcast episode?
Free for the download + Whisper transcription + clip detection. The AI virality scoring step uses cloud credits (~$0.02 per candidate clip evaluated). A 60-minute episode that surfaces 15 candidate clips costs roughly $0.30 total in cloud credits. No monthly subscription, no per-minute transcription cap.
Does it work for video podcasts (Joe Rogan style) too?
Yes — and that's actually the cleanest fit. Drop the YouTube video podcast URL, ViralMint downloads the full video locally, transcribes, scores clips, then exports each pick as a vertical 9:16 short with face-tracking (so the speaker stays centered when the original 16:9 is reframed) and word-by-word captions burned in. Audio-only podcasts work the same way but you'd want to add a static cover-art video underneath the audio track for the visual.
How is this different from Opus Clip's podcast workflow?
Opus Clip is a hosted cloud service that uploads your podcast for processing and charges monthly. ViralMint runs everything locally — your podcast audio never leaves your machine — and charges per scored clip instead of monthly. Feature parity on the core extract-and-caption flow; ViralMint adds local-first privacy + no monthly minimum + open-source code, while Opus Clip wins on browser-first UX and direct YouTube auto-upload.
Get ViralMint
The Clip Extractor + Whisper transcription ship inside the free desktop app — no subscription, no monthly cap, no watermark. Free for the download, prepaid credits for the AI virality scoring (~$0.30 per 60-min episode).