Xiaohongshu (RedNote) Video Transcript Scraper

📕 Xiaohongshu (RedNote) Video Transcript Scraper

Turn Xiaohongshu (RedNote) videos into searchable text. Pick a category, and the Xiaohongshu Video Transcript Scraper returns the platform's trending video posts with an AI transcript of every one alongside the full post data — title, caption, hashtags, author, likes, saves, comments and shares.

RedNote is where Chinese consumers actually discuss products, and roughly half of what the platform surfaces is video. None of it carries captions, and none of it is indexed in a form you can read, search, or feed to a model. This Actor fixes that — it makes China's most commercially interesting social platform legible to someone who doesn't read Chinese.

Run it on the Apify platform to get API access, scheduling, integrations (Google Sheets, Make, Zapier, and more), automatic retries, proxy rotation, and monitoring out of the box — no code required.

✨ Features

• 🎙️ AI Transcription: Every video turned into clean text, tuned for Mandarin and accurate on technical and product vocabulary.
• ⏱️ Timestamped Segments: Transcripts come split into properly bounded segments with start and end times — ready for clipping, subtitling, or jumping to a quote.
• 🗂️ 12 Topical Categories: Sample trending video from Fashion, Food, Beauty, Travel, Fitness, Gaming, Movies & TV, Career and more.
• ♻️ Fresh Results Every Run: Each run pulls a new slice of the feed, so a daily schedule keeps building your archive instead of repeating itself.
• 🔇 No Empty Transcripts: About half of RedNote's videos are music over footage with nobody talking. Those are detected and filtered out up front, so your results are videos that actually have something to say.
• 📈 Richer Engagement Than Western Platforms: Likes, saves, comments and shares on every record — saves being the metric that actually signals purchase intent on RedNote.
• 🏷️ Hashtags and Location: The post's full topic list and, where the author exposes it, the region they posted from.
• 🎬 Direct Media Links: The signed video URL plus its unsigned mirrors, so a link that expires is still recoverable.
• ⏳ Length Guard: A maximum video length, so one hours-long livestream replay can't blow up a run's cost.
• 💸 Never Billed Twice: The same video appearing under several posts is transcribed once and charged once.

🛠️ How to Use the Xiaohongshu (RedNote) Video Transcript Scraper

1. Pick your categories – Choose one or more, such as Beauty & cosmetics or Food. Leave it on Video for the platform's general video feed.
2. Set your options – Choose how many posts to return, a maximum video length, and the spoken language (Mandarin by default).
3. Run the scraper – Click Start. Image-only posts and silent videos are filtered out before they cost you anything.
4. Export your data – Download the dataset as JSON, CSV, Excel, or HTML, or pull it via the Apify API.

Already know which posts you want? Paste their URLs into Post URLs instead — copy them from the site with the full query string intact.

🔧 Input

Field	Type	Description
categories	Array	Categories to sample trending video from. 12 available, from Fashion to Career.
postUrls	Array	Specific post URLs, copied from the site with their complete query string.
onlyVideosWithSpeech	Boolean	Return only videos that actually contain speech. On by default.
maxResults	Integer	Maximum number of video posts to return (default 20).
maxVideoMinutes	Integer	Videos longer than this are returned without a transcript, and not charged for (default 20).
language	Select	Spoken language of the audio. Defaults to Chinese; auto detects it instead.
proxyConfiguration	Object	Proxy settings. Keep this enabled — the source limits how much a single address may request.

📊 Sample Output Data

[
    {
        "noteId": "6a55886f00000000150241a2",
        "postUrl": "https://www.xiaohongshu.com/explore/6a55886f00000000150241a2?xsec_token=AB4G...",
        "postedAt": "2026-07-12T09:33:04.000Z",
        "discoveredVia": "homefeed.food_v",
        "title": "辛拉面到底怎么煮才好吃",
        "description": "#峰哥[话题]# #辛拉面[话题]# #方便面[话题]# 今天来试试这个做法…",
        "tags": ["峰哥", "辛拉面", "方便面"],
        "authorName": "峰哥正能量",
        "authorId": "650b894e00000000120070c9",
        "authorUrl": "https://www.xiaohongshu.com/user/profile/650b894e00000000120070c9",
        "ipLocation": "上海",
        "likeCount": 2084,
        "collectedCount": 378,
        "commentCount": 1281,
        "shareCount": 597,
        "durationSeconds": 144,
        "videoUrl": "https://sns-video-v28.xhscdn.com/stream/1/110/258/....mp4?sign=...&t=...",
        "videoBackupUrls": ["http://sns-bak-v8.xhscdn.com/stream/1/110/258/....mp4"],
        "videoWidth": 720,
        "videoHeight": 1280,
        "thumbnailUrl": "https://sns-webpic-qc.xhscdn.com/....webp",
        "subtitles": [],
        "transcript": "大家好今天我们来聊一聊这个辛拉面的正确煮法…",
        "transcriptLanguage": "zh",
        "transcriptConfidence": 0.997,
        "transcriptSegments": [{ "text": "大家好今天我们来聊一聊…", "start": 0.1, "end": 12.4, "confidence": 0.99 }]
    }
]

You can download the dataset in various formats such as JSON, HTML, CSV, or Excel.

📋 Data Fields

Field	Description
noteId / postUrl	Post identifier and direct link.
postedAt	When the post was published.
discoveredVia	Which category the post was found in, or direct for a pasted URL.
title, description, tags	The post's headline, caption text, and hashtag topics.
authorName, authorId, authorUrl	Who posted it, and a link to their profile.
ipLocation	Region the author posted from, where they expose it.
likeCount, collectedCount, commentCount, shareCount	Engagement — likes, saves, comments and shares.
durationSeconds	Video length in seconds.
videoUrl, videoBackupUrls	Direct media link, plus mirrors that outlive the signed URL.
videoWidth, videoHeight, thumbnailUrl	Video dimensions and cover image.
transcript	The full spoken text of the video.
transcriptLanguage, transcriptConfidence	Language and the model's confidence score.
transcriptSegments	Timestamped segments with text, start, end, and confidence.
subtitles	Platform-generated subtitle tracks, on the rare posts that have them.
transcriptNote	Present when a transcript was skipped, and why.

💵 How Much Does It Cost to Scrape Xiaohongshu Videos?

This Actor uses the pay-per-event pricing model — you only pay for what you get:

Event	Price	When it's charged
Actor start	$0.00005	Once per run.
Post scraped	$0.006	Per post returned, with all its metadata.
Video transcribed	$0.01	Once per video turned into text.
Minute of video transcribed	$0.01	Per started minute of audio.

Half of the videos on this platform run under 65 seconds, so a typical transcribed post costs about $0.036 — roughly $3.60 for 100 videos. Longer videos cost proportionally more, which is exactly why the Max video length option exists and defaults to 20 minutes. Image-only posts are filtered out before they cost anything, videos containing no speech are not charged for, and a video appearing under several posts is charged once. New Apify accounts include free monthly usage credits, so you can try it at no cost.

💡 Tips

• Start with a small maxResults to preview the data shape before scaling up.
• Feed the transcripts into an LLM to translate, summarise, or pull product mentions at scale — the transcripts come out as continuous Mandarin text, which is ideal input for a model.
• Schedule a daily run on the categories you care about. Each run returns a fresh slice, so a week of runs gives you a genuine trend picture rather than the same posts again.
• Sort your dataset by collectedCount rather than likeCount — saves are the strongest intent signal on this platform.
• Lower Max video length to 5 or 10 minutes when you only want short-form content and want costs tight.

⚠️ Good to Know

• This Actor samples what the platform is currently promoting in each category. It is not a keyword search and not a creator monitor — you choose a category, not a search term or an account.
• The feed serves a different set on every request, which is what makes repeat runs valuable, but also means results are not reproducible run to run by design.
• Roughly half of the posts on the platform are image-only. Those are detected and skipped before any request is spent on them.
• Of the videos, roughly half again carry no speech — short clips set to music. Only videos with speech is on by default so those don't fill your dataset; turn it off if you want them as plain post records.
• A small number of posts are hours-long livestream replays. The Max video length guard returns their post data without transcribing them.

❓ FAQ

Do I need an account or cookies? No. The Actor only reads publicly visible posts — no login, no cookies, nothing to configure.

Do I need a proxy? Yes, and it's on by default. The source stops serving results to an address that has requested too much, so proxy rotation is what keeps a longer run working.

Can I search by keyword or scrape a specific creator? Not with this Actor. Those surfaces are not publicly readable on this platform; category sampling and direct post URLs are.

How accurate are the Mandarin transcripts? Very. Measured confidence sits around 0.997 on clear speech, and each record carries its own transcriptConfidence so you can filter on quality. Note that Chinese transcripts come back with little punctuation — that's a property of the language model output, and it doesn't affect keyword search or LLM processing.

Is scraping Xiaohongshu legal? This Actor collects only publicly available post data. You are responsible for using the data in compliance with Xiaohongshu's Terms of Service and applicable laws.

What about personal data? Records include the post author's public nickname, user ID and profile URL, all already public. If you're processing this data for EU residents, you're the controller and GDPR applies to what you do with it.

🆘 Support

Found a bug or need a custom data field? Open an issue from the Actor's Issues tab and we'll take a look. Custom scraping solutions are also available on request.

Make RedNote's video content searchable with the Xiaohongshu (RedNote) Video Transcript Scraper today! 🚀