I Tested 6 Ways to Extract Locations from Travel Reels
I have a problem. Or — okay, fine — most of us have this problem. My Instagram saves folder has 340-something travel Reels. Bali beaches. Tokyo ramen. That one Reel about a "secret" pasta restaurant in Rome that probably is not secret. None of the places are written down anywhere. They're just stuck inside the videos.
For two weeks I tried every method I could find for extracting place names out of those Reels. Some I thought would be silly and they sort of worked. Some I assumed would be magical and they were a slog. I tested six total. Two are great. Two are tolerable. Two waste your time.
This is the honest version. No "here's a tip and now buy something" energy. Just what happened when I tried it.
The Test Setup
I picked 30 saved Reels off my profile, picked at random — short ones, long ones, ones with one place, ones with twenty. The benchmark for "did it work" was simple: at the end, do I have a list of named places I can search in Google Maps, with the right spelling, and can I tell which Reel they came from?
Here's what I tried.
1. AI Extraction (Triply) — The One That Just Works
Share the Reel into Triply. Wait 10–15 seconds. Done. Every place named in the Reel — audio, captions, on-screen text — comes back as a mapped list. The first time I did this with a "12 best food spots in Bangkok" Reel, I just sat there. It got 11 of 12. The 12th was a stall the creator only pointed to without naming.
Across the 30 Reels in my test, this caught about 92% of named places. The misses were almost all visually-only spots (creator points camera, says "this place is amazing" without naming it). For anything where the place is said or written, it nailed it.
- ~10 sec per Reel — by far the fastest
- 92% extraction rate across my 30-Reel test
- Catches spoken names AND on-screen captions
- Geocodes everything automatically
- Outputs to a map, not just a list
- iOS only (annoying for Android friends)
- Free tier caps monthly extractions
- Visually-only places get missed
Verdict: Top pick. Two weeks in, this is the one I actually kept using.
2. Manual Rewatch + Google Maps Search — Free, Slow, Honest
The OG method. Rewatch each Reel, write down place names, search them in Google Maps, save to a list. Free. Works. Tedious in a way that makes you question your hobbies.
Average time per Reel for me: 2 minutes 40 seconds. Multi-place roundup Reels routinely took 5+ minutes. Extraction quality was about 78% — I'd skip places I didn't catch the first time, mishear names (RIP "Yamashita" turning into "Yamasa"), and occasionally get the wrong restaurant in the search.
- Free, no apps
- End result is real Google Maps lists
- Forces you to actually engage with each Reel
- ~2:40 per Reel — extrapolate that to 100 Reels
- 78% extraction; you'll mishear or miss places
- Tedious in a way that breaks your will
Verdict: Fine for 5 Reels. For 50? You will lose the will to live.
3. Whisper Transcription Pipeline — Cool If You're a Nerd
I downloaded each Reel as MP4, ran them through OpenAI Whisper for transcription, then fed the transcripts into GPT-4 with a prompt like "list every place name." Then geocoded with the Google Maps API. I built this in an afternoon because I was curious.
Result: works, kinda. Whisper struggled when Reels had loud background music (lots of them do). GPT-4 was decent at parsing transcripts but missed any place that was only shown on screen as a caption (lots of Reels do that too). Total extraction across 30 Reels: about 71%. Setup time: ~3 hours. Per-Reel processing: ~30 sec. API costs: roughly $7.
- Once built, fast per Reel
- Full control over the pipeline
- Fun afternoon project if you're into this stuff
- 3 hours of setup before you process a single Reel
- 71% extraction — Whisper hates music, GPT misses captions
- API costs creep up at scale
- You still need to geocode and map separately
Verdict: A worse version of Method 1, but more configurable. Worth it only if you genuinely want to tinker.
4. DM the Creator and Ask for the Address — The Hopeful Hack
I'd seen people swear by this. "Just DM the creator! They always reply!" I DM'd 12 creators across 12 Reels. Three replied. Two of those gave actual addresses. The other one sent me a link to their TikTok shop selling matcha powder.
Math: 12 messages → 2 useful answers → 16% hit rate, multi-day turnaround. To extract places from 30 Reels at this rate would have taken me weeks and gotten me five answers. This is not a method. This is a coping mechanism.
- When it works, you get the actual creator's spot
- Builds a parasocial bond, I guess
- 16% reply rate, multi-day delays
- Doesn't scale past ~5 Reels without becoming a part-time job
- Most replies aren't usable
Verdict: Touching that this still works at all. Useless as a real method.
5. Browser Extensions That Promise to "Extract Reels Data" — Mostly Junk
I tried four of these. They promise to scrape Reel comments, captions, and metadata. None of them actually extract place names from the video itself. They scrape what's already text — captions and comments. So if a Reel has a caption with addresses, you get the addresses. If it doesn't (most don't), you get nothing.
Across my 30 Reels, this approach caught maybe 15% of places — only the ones helpfully listed in the caption. Most travel Reels don't list addresses in the caption. The whole appeal of these tools is "AI extraction" but none of them are doing AI extraction; they're just CSV-dumping the visible text.
- Sometimes pulls usable info from caption-heavy Reels
- Free or freemium
- ~15% extraction across my test
- None of them analyse the video itself
- Many require Instagram login (sketchy)
- Misleading marketing — "AI extraction" usually means "we copy the caption"
Verdict: Skip. The marketing is more impressive than the product.
6. Screenshot OCR (iOS Live Text / Google Lens) — Surprisingly Decent for Captions
This one I expected to be silly and it ended up being legitimately useful — for one specific kind of Reel. iOS Live Text and Google Lens can OCR text out of any screenshot. So if a Reel shows a caption like "Bar Pasalacqua, Lombardia," you screenshot, tap-and-hold the text, copy, paste into Google Maps. Done.
For Reels that rely on bold on-screen text, this works. Across my 30 Reels, about 60% had decipherable on-screen text and OCR caught those well. The other 40% (audio-only, no text) gave OCR nothing. Per-Reel: maybe 30 seconds if there are 3-4 places shown.
- Genuinely fast for caption-heavy Reels
- No app install — uses tools you already have
- Free
- Useless for audio-only Reels (~40% of mine)
- Requires you to manually screenshot every place — still tedious
- You still have to search and save each one in Maps yourself
Verdict: A reasonable backup tool for captions, especially in combination with another method. Not a complete solution on its own.
The Scoreboard
Across my 30-Reel test, ranked by combined extraction quality and time per Reel:
- 1. Triply (AI extraction) — 92% extraction · ~10 sec/Reel · the clear winner
- 2. iOS Live Text / Lens OCR — 60% extraction (caption-only) · ~30 sec/Reel · solid backup
- 3. Manual rewatch — 78% extraction · ~2:40/Reel · works but kills your weekend
- 4. Whisper + GPT pipeline — 71% extraction · ~30 sec/Reel after 3 hr setup · for tinkerers only
- 5. Browser extension scrapers — 15% extraction · varies · mostly skip
- 6. DM the creator — 16% reply rate · days · charming, not useful
What I Actually Do Now
Here's the workflow I landed on after two weeks of testing: Triply for everything, with iOS Live Text as a fallback for the rare Reel where extraction misses something obvious in the caption.
I haven't manually rewatched a Reel in three weeks. The 340-Reel saved folder is now slowly becoming a real map. I'm catching up on it during commutes — share, wait 12 seconds, drop into a collection, repeat. The whole "I'll plan a trip from these someday" thing is finally happening, and the someday is in three weekends.
If you only take one thing from this: the gap between AI extraction and every other method is enormous. Manual works. Whisper works. OCR works for captions. But the time-to-mapped-place metric is in a different universe with proper extraction. Once you've experienced the 10-second flow, you can't go back to two-minute rewatches.
Frequently Asked Questions
How do you extract a location from an Instagram Reel?
Fastest: share the Reel into an AI extraction app like Triply. It analyses the audio, captions, and on-screen text and returns every place mentioned, geocoded onto a map. Slower alternatives include rewatching and searching each place in Google Maps, using OCR on on-screen captions, or DM-ing the creator.
Can ChatGPT find locations in Instagram videos?
Not directly — ChatGPT can't watch videos. You'd need to first transcribe the audio (Whisper) and pass the transcript plus any captions into ChatGPT for parsing. That works for audio-clear Reels but misses any place shown only on screen. Purpose-built tools handle this whole pipeline natively.
Is there a Chrome extension to extract locations from Instagram Reels?
There are caption and comment scrapers, but they don't analyse the video content itself — so they miss most travel places. For real video-content extraction you need a purpose-built mobile app like Triply.
How accurate is AI location extraction from short videos?
In my two-week test, AI extraction caught about 92% of named places across 30 Reels. Misses were almost all visually-only spots (creator shows but doesn't name them). For audibly or textually named places, accuracy is very high.
What's the best free way to find addresses from travel Reels?
The best free way is rewatching the Reel and searching each named place in Google Maps. It's slow (~2:40 per Reel in my test) but free. For larger collections, Triply's free tier is faster even with its monthly cap.
Skip the experiments. Just extract the places.
Triply turns a saved Reel into a mapped trip in seconds — no setup, no scripts.
Get Triply