There's a specific kind of regret that hits around video #300. You look back at years of uploads — interviews, tutorials, vlogs — and realize none of them have transcripts. Old videos with transcripts surface in search for long-tail queries. Old videos without them are invisible.
On Hacker News, the requests are blunt: someone asking to "convert a whole playlist in bulk" (hn_38236198), and someone describing the extreme case — "1,800 video diaries... don't have transcripts" (hn_34039362). Eighteen hundred. Even at a generous 30 seconds of handling per video, that's 15 hours of pure clicking before any transcription happens.
Bulk is a different problem from single-video transcription. Here's how to think about it.
Why back-catalog transcription pays
Search engines index text, not speech. YouTube auto-captions help somewhat, but they're error-prone, not downloadable as clean text, and not under your control. A corrected transcript on your own site — or a corrected caption file on the video — turns each old video into a page that can rank.
The math for a 300-video catalog: if transcripts add even 5-10 long-tail search visits per video per month, that's 1,500-3,000 incremental views monthly from content you already made. The work is one-time; the traffic compounds.
The catch is the word "one-time" — at scale, it's still 300 × (download + transcribe + correct + format + upload). Every step multiplied.
The options at scale
One video at a time via web tools. Most transcription SaaS is built for the single-upload workflow. Doing it 300 times means 300 uploads, 300 waits, 300 downloads — and at $12-29/month subscription tiers with per-minute caps, the pricing is designed against bulk.
Agencies and human transcription. At $1-3/audio minute, a 300-video catalog averaging 12 minutes each runs $3,600-$10,800. Non-starter for most creators.
YouTube auto-captions, en masse. Free and already there. Quality varies from okay to disgraceful, they live only on YouTube, and you can't cleanly export corrected versions.
The DIY path: a real bulk pipeline
For technical users, this is a solved problem if you write ~20 lines of glue:
-
Fetch the playlist.
yt-dlp --flat-playlist -J "<playlist_url>"lists every video ID. For your own channel, the uploads playlist works the same way. -
Download audio only, in a loop.
yt-dlp -x --audio-format mp3 -o "%(id)s.%(ext)s" "<playlist_url>". Audio-only downloads are small — a 15-minute video is ~10-15 MB — so 300 videos fit on a modest drive. - Batch transcribe. Locally with Whisper large-v3 (slow on CPU: budget 1-2x real-time per video, so a full day of compute for a big catalog) or via an API like Groq's Whisper endpoint, which processes a 15-minute file in well under a minute and costs roughly a cent. A 300-video catalog via API: a few dollars.
-
Generate SRT files (
whisper --output_format srt) and upload them to the corresponding videos — YouTube Studio allows caption upload per video, and there's an API for it if you're doing hundreds. - The correction problem. Here's where bulk DIY cracks. Step 3's output has the same proper-noun errors as single-video transcription, multiplied by 300. Nobody corrects 300 transcripts. In practice, DIY bulk means choosing between shipping known-wrong captions or never finishing.
That last step is the actual hard part of back-catalog work, and it's the one most tooling ignores.
How postwriter.cn handles bulk
Bulk mode is one of the three things postwriter.cn was built around. Paste a playlist URL — or upload a folder of files for off-YouTube content — and the whole thing queues. Each video comes back with transcript, SRT, description, chapters, titles, and social copy. You're reviewing outputs, not babysitting uploads.
The personal dictionary does the heavy lifting at scale. Correct "Marques" or "framework" once, and correction applies across the batch — including the videos you haven't reviewed yet, since later runs draw on everything learned from earlier ones. For the 1,800-video-diary extreme, that convergence is the difference between a catalog that gets more accurate as you process it and a catalog that's uniformly 85% accurate forever.
Pricing is built for bulk economics too: free during beta, $39 for 3 years at founder pricing. Compare a $19/month tool over three years — $684 — for the same batch capability with per-minute caps, and the arithmetic is blunt.
FAQ
Can I transcribe videos that aren't mine?
Technically yes for personal research use; the copyright rules for republishing transcripts of others' content are restrictive. For your own catalog, it's your content — no issue.
Does adding captions to old videos actually improve their performance?
Yes, typically. Caption files add indexable text, and old videos with new metadata often see search rediscovery within weeks. It's one of the highest-ROI actions available on a back catalog.
How long does bulk transcription take?
With API-based processing, roughly real-time for the whole batch divided by concurrency — a 40-hour catalog can process in under an hour of wall time. Local CPU Whisper is far slower; GPU changes that. The human review pass is usually the bottleneck, which is why dictionary-based convergence matters at scale.
What about podcast back catalogs?
Same pipeline, audio-only. RSS feeds make fetching even easier than YouTube playlists, and the transcripts feed show notes and blog repurposing directly.
Is $39/3 years sustainable pricing or a teaser?
It's a founder-tier price for early users, stated plainly: the first cohort pays $39, locking three years, and the point is to price against the $12-29/month standard rather than match it. Whether the product earns renewal in year four is the bet we're making.
Top comments (0)