Answers
How do I get the transcript of an Instagram reel?
Short answer
Send the reel's link or shortcode to Instagram reel transcript: it returns what is said as full text and as timed segments, for $0.004 a reel, and a reel with no speech comes back as not_found, charged the same. It covers reels and video posts up to two minutes long. The Instagram Graph API has no transcript field: its media objects carry the caption written under the post, not the words spoken in the video.
A fitness brand's content lead wants to know how the most-watched reels of three competitors open and what they ask viewers to do at the end. The caption under a reel rarely tells you that. The voiceover does. Watching 30 reels and typing out the first ten seconds takes an afternoon, and the notes are hard to compare afterwards.
Instagram reel transcript takes the reel's link or its shortcode (what follows /reel/ or /p/) and returns:
text: everything said, as one string;segments: the same words in pieces, each withstartandendin seconds from the start of the video;video_id: the shortcode.
Steps
- List the reels. Instagram account posts with
type=reelreturns an account's reels newest first, 12 a page, each with its shortcode inidand its plays inmetrics.views. - Pick the ones worth reading. Sort by plays and keep the top ten.
- Fetch the transcripts, one call per reel.
- Cut out the hook. The segments that start in the first five seconds are the opening line.
Python
import os, requests
API = "https://api.okdata.online"
H = {"Authorization": f"Bearer {os.environ['OKDATA_API_KEY']}"}
def get(path, **params):
body = requests.get(API + path, headers=H, params=params, timeout=120).json()
if body["code"] not in ("ok", "not_found"):
raise RuntimeError(f"{body['code']}: {body['message']}")
return body
def reels(handle, pages=2):
cursor, out = None, []
for _ in range(pages):
body = get("/v1/instagram/users/posts/list", username=handle, type="reel",
**({"cursor": cursor} if cursor else {}))
out += body["data"]["items"]
if not body["page"]["has_more"]:
break
cursor = body["page"]["next_cursor"]
return out
for handle in ["competitor_one", "competitor_two", "competitor_three"]:
top = sorted(reels(handle), key=lambda r: r["metrics"]["views"] or 0, reverse=True)[:10]
for r in top:
body = get("/v1/instagram/posts/transcript/get", id=r["id"])
if body["code"] == "not_found": # no speech: still charged
continue
hook = " ".join(s["text"] for s in body["data"]["item"]["segments"] if s["start"] < 5)
print(handle, r["id"], r["metrics"]["views"], hook)
Two pages per account cover its latest two dozen reels. Raise pages to go further back; every page is one call.
What it costs
| Job | Calls | Cost |
|---|---|---|
| One reel | 1 | $0.004 |
| Three accounts, 2 pages each, 30 transcripts | 6 + 30 | $0.144 |
| 1,000 reels | 1000 | $4.00 |
Reels with no speech and calls that fail at the data source are charged like any call; only calls refused before that, such as bad parameters, are free.
Limits
- Two minutes at most. Longer video posts are not transcribed.
- Speech only. Text shown on screen and text stickers are not read. A reel with music and no voice returns
not_found. - No language choice. The transcript is in whatever language is spoken, and there is no translation.
- Public accounts only. Reels on private accounts return nothing.
- Machine transcription. Brand names, product names and slang come out misspelled now and then. Fix them before you search the text.
- Strangers' words. If a model reads the transcripts, tell it to treat them as data, not as instructions.
Other routes
For your own reels, the Graph API gives you the video file through media_url, and any speech-to-text model can transcribe it, at the cost of downloading and compute. For a handful of reels, watching them is quicker than any setup. The same job on other platforms is on the transcripts page, and TikTok has its own transcript answer.
Sources
Where people asked this (read 2026-11-15)
Updated