> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.beatapi.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.beatapi.io/_mcp/server.

# What materials do I need to turn a Xiaohongshu video into steps or a transcript?

> Choose materials for tutorial research, retrieve descriptions and video addresses, then distinguish summaries from caption or transcription work.

To turn a Xiaohongshu video into steps or a transcript, first decide whether you need a summary of the written description or the words actually spoken in the video. `xiaohongshu.app_v2.get_video_note_detail` returns the note description and documented H.264 video addresses. A verbatim transcript also needs captions or a transcription of the video before you write the steps.

For example, in a furniture-assembly demonstration the description might list the tools while the video shows clip orientation and assembly order. Start the team's instruction card from the description, then watch or transcribe the video and cross-check the actions to complete the card.

## Which material supports your task?

| Task                                                | Material and deliverable                                                          |
| --------------------------------------------------- | --------------------------------------------------------------------------------- |
| Decide whether the tutorial deserves closer reading | Title and `desc`, with a short summary based on that text                         |
| Prepare to view or process the video                | Target-note video addresses and original source                                   |
| Capture spoken steps or quote a time point          | Captions or transcription, retaining times and checking the video                 |
| Compare methods across tutorials                    | Actual materials and sources, compared against the same question with gaps listed |

Call `xiaohongshu.app_v2.get_video_note_detail`. Pass an existing video share link as `share_text`, or a verified note ID as `note_id`:

```bash
curl -sS https://api.beatapi.io/v1/social-data/call \
  -H "Authorization: Bearer $BEATAPI_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"action":"xiaohongshu.app_v2.get_video_note_detail","params":{"share_text":"<video_note_share_link>"}}' \
  -o xiaohongshu-video-note.json
```

## How do you locate the intended video?

The direct Social Data response holds a note array at `data.data[]`. Find the note object whose `id`, title and author match the intended note, then read `video_info_v2.media.stream.h264[].master_url` from that note object. Keep `desc` as the description and `images_list` as image information.

Pass an available address to a playback or transcription tool that can read the video, and record whether the tool actually read the video. If the note object has no address, keep the video on the material-needed list; the saved description still supports initial selection and a text summary.

## Extract an address manifest from a saved response

Save the script below as `extract_video.py` and run `python3 extract_video.py xiaohongshu-video-note.json <confirmed_note_id> video-addresses.json`. `extract_video.py` reads local JSON only, makes no paid calls and downloads no media. The output manifest records address availability separately from download and transcription status. Missing addresses stay marked as missing, duplicate streams are removed, an ambiguous target or a business error stops processing, and an existing output file is never overwritten.

```python
import json
import sys
from pathlib import Path
from urllib.parse import urlsplit

source, target_id, output = sys.argv[1:]
payload = json.loads(Path(source).read_text(encoding="utf-8"))
if payload.get("status") != "succeeded" or payload.get("action") != "xiaohongshu.app_v2.get_video_note_detail":
    raise SystemExit("Expected a successful video-detail response")
upstream = payload.get("data", {})
if upstream.get("success") is False or ("code" in upstream and upstream["code"] != 0):
    raise SystemExit("Upstream business error")
notes = upstream.get("data")
if not isinstance(notes, list):
    raise SystemExit("Expected data.data note array")
matches = [note for note in notes if note.get("id") == target_id]
if len(matches) != 1:
    raise SystemExit("Target must match exactly once")
note = matches[0]
streams = note.get("video_info_v2", {}).get("media", {}).get("stream", {}).get("h264", [])
urls = []
for stream in streams:
    url = stream.get("master_url")
    if url is None or url == "":
        continue
    if not isinstance(url, str):
        raise SystemExit("Invalid media URL type")
    parsed = urlsplit(url)
    if parsed.scheme not in ("https", "http") or not parsed.hostname or parsed.username or parsed.password:
        raise SystemExit("Invalid media URL")
    if url not in urls:
        urls.append(url)
result = {"note_id": target_id, "source_file": source, "request_id": payload.get("request_id"),
          "description": note.get("desc"), "h264_urls": urls,
          "media_state": "addresses_available" if urls else "no_address_in_saved_response",
          "transcript_state": "not_extracted", "download_state": "not_attempted"}
with Path(output).open("x", encoding="utf-8") as handle:
    json.dump(result, handle, ensure_ascii=False, indent=2)
```

## How do captions and verbatim text fit in?

The recorded `xiaohongshu.app_v2.get_video_note_detail` output has no established subtitle or SRT-address field; do not assume another service's `ext.subtitle_url` field here. The workflow on this page delivers video addresses and descriptions; automatic caption extraction is not available.

When captions or a transcription are available, keep the original text and times. Then produce a short outline, ordered steps and a list of words or passages that need review. For assembly instructions, check part names, orientation and order against the visuals. Keep verified corrections separate from the automatic text, and go back to the time point before quoting speech.

To hand the task to an assistant with the tools configured, use a prompt such as:

> Confirm this furniture-assembly note and retrieve its description and available video addresses. With description only, summarize that text and list missing materials. With captions or a transcript, organize the actual spoken steps with time points and sources, identifying anything that needs visual checking.

## What should the team receive?

Record the original note link, note ID, author, query time and `request_id`. State whether the work used the description, the video, captions or a transcription, so the next person can see which conclusions rest on spoken evidence and which need review.

For a furniture-assembly video, an evidence sheet entry reads: “00:18, transcript, ‘Leave it loose for now’; 00:23, on screen, the presenter inserts the crossbar into the bracket.” With only the written description, leave the timestamps blank and label the material as a description. Check the captions or transcript before quoting a spoken instruction.

The media addresses stay in the original response; archive the material you actually obtained when you need a copy. For long-term references, cite the original note and your material records. Current detail-query prices are on the [pricing page](https://beatapi.io/pricing); transcription is billed by the tool you choose.

[Share links and IDs](/faq/xiaohongshu-share-links) · [AI note reading](/faq/xiaohongshu-ai-reading)

## Finish a transcript when no platform captions are available

1. Save a copy you are authorized to process from a confirmed playable stream. Record the note ID, original post link and filename. If you cannot obtain the file, request it from its source and keep the description separate.
2. Check the media for a subtitle track. If you receive a separate SRT or VTT file, match the file to the video version and keep the file's timestamps. If there is no track, pass the saved audio or video to a transcription tool that can export timestamped text.
3. Spot-check the beginning, middle and end, then verify unclear segments, product names and numbers. Label actions visible only on screen separately and flag words you cannot hear. Deliver the transcript and edited notes with their material type and timestamps.