Krakt Threadtrack
Pricing
$20.00 / 1,000 thread processeds
Krakt Threadtrack
Paste a Reddit, Hacker News or Bluesky thread. Get every song people mentioned as a clean, deduped tracklist ranked by mentions, verified on MusicBrainz, with Spotify and YouTube links plus a one-click YouTube playlist. About $0.02 per thread.
Pricing
$20.00 / 1,000 thread processeds
Rating
0.0
(0)
Developer
Kurt Holes
Maintained by CommunityActor stats
0
Bookmarked
2
Total users
1
Monthly active users
8 days ago
Last modified
Categories
Share
Somebody asked "what song gave you chills?" and 400 people answered. This actor turns that thread into a playlist.
Paste a Reddit, Hacker News or Bluesky thread. Krakt Threadtrack reads every comment, pulls out every song people mentioned, checks each one against MusicBrainz, merges the duplicates and ranks them by how many different people brought them up. You get a clean tracklist with Spotify and YouTube links and a one-click YouTube playlist. No AI, no login needed.
What you get
| field | what |
|---|---|
rank, mentions | Ranked by how many different people mentioned the song |
artist, title | Cleaned up and spelled the MusicBrainz way |
sampleComments | Up to 3 quotes (one per person) that mention it |
spotifySearchUrl, youtubeMusicSearchUrl, youtubeSearchUrl | One click to play it anywhere |
youtubeUrl, spotifyUrl | Direct links when we found the exact track |
youtubeQuickPlaylistUrl | One link that plays the top 50 as a YouTube queue |
spotifyPlaylistUrl | A real private Spotify playlist, if you add your own Spotify token |
verified, musicbrainzId, year, foundVia | Extras |
The per-thread summary (comment count, track count, quick playlist links) is in the OUTPUT record of the key-value store.
How songs are found
Artist - Titlelines,"Title" by Artist,Title by Artist,Artist's "Title"- Spotify track links and YouTube / YouTube Music links in the comments (read from the public page and oEmbed)
- Every guess is checked on MusicBrainz (free, 1 request/second, proper User-Agent). Guesses it can't confirm are dropped unless they came from a link.
Use it for
- "Songs that…" threads → instant playlist
- Music newsletters, radio prep, DJ crate digging
- Finding what a community actually loves, ranked
Pricing
$0.02 per thread. Threads where no songs are found are free.
Optional: real Spotify playlist
Paste a Spotify access token with the playlist-modify-private scope into Spotify access token. The actor creates a private playlist in your account and adds every track it can match. The token is only used for that run and is stored as a secret input.
Notes and limits
- Reddit: Reddit blocks cloud servers from its JSON API (
old.reddit.com/.json,api.reddit.com). The actor tries those, then Reddit's public comment RSS feed (works today, up to ~500 comments per thread, no vote counts), then the same routes through Apify's datacenter proxy. If Reddit tightens up, Reddit threads may fail; HN and Bluesky are rock solid. - Hacker News (Algolia API) and Bluesky (public AppView) read the full thread.
- X/Twitter is not supported. Its API is paid.
- MusicBrainz allows 1 request/second, so verification is time-boxed (45 seconds per thread by default); the most-mentioned candidates are checked first. For a big thread, raise MusicBrainz time budget (e.g. 240). If nothing could be confirmed in time, the unconfirmed guesses are returned with
verified: falseinstead of an empty list. - The whole run has a time budget (Run time budget, 240 seconds by default). When it runs out, results found so far are saved and remaining threads are skipped (not charged). Raise it, together with the run timeout, for many threads.
- Text rules miss songs written as bare titles with no artist ("Wonderwall!") and sometimes pick up albums instead of songs.
- The YouTube quick playlist uses YouTube's
watch_videoslink with video IDs looked up from public search results. If lookups fail, you still get search links. - Spotify has no login-free way to make a combined playlist link, so without a token you get per-track Spotify search links.
- Comments belong to their authors. Link back when you share.