Comments Extractor — YouTube + Reddit, One Schema (+AI)

Comments Extractor — YouTube + Reddit, One Schema (+AI)

Extract comments from YouTube videos and Reddit posts into a single normalized schema — author, text, likes/score, reply count, posted date, and URLs — so you don't have to write separate parsers for each platform. An optional AI pass tags every comment with sentiment and an intent flag (spam/lead/complaint).

What it does

Give the actor a list of YouTube video URLs and/or Reddit post URLs. For each one it pulls the top-level comments (YouTube via its comment endpoint, Reddit via the Arctic Shift API) and maps them into one common row shape, so a downstream pipeline that consumes "comments" doesn't need platform-specific branching. Raw comment extraction is the core product; AI tagging is an add-on you switch on per run, not a separate actor.

Input

• YouTube video URLs (youtubeUrls) — full URLs or bare 11-character video IDs.
• Reddit post URLs (redditUrls) — post URLs in the …/comments/<id>/… form.
• Max comments per video/post (maxCommentsPerContent) — how many comments to pull from each item (default 50, 1–1000). You are charged per comment actually returned.
• AI: tag sentiment + intent (enrichComments) — boolean, off by default. When on, each comment gets a sentiment (positive/neutral/negative) and a flag (spam/lead/complaint/none).
• Your OpenAI API key (openaiApiKey) — optional, secret field. If you supply your own OpenAI-compatible key, the AI pass runs on your account and you pay your provider directly. Leave it empty and the actor still tags every comment: it falls back to a built-in classifier (a bundled key if configured, otherwise a free deterministic lexicon classifier) — enrichment always returns a value, it's never blocked on a missing key.
• AI model (aiModel) — optional override of the model used for AI tagging (default gpt-4o-mini); only applies when a key is available.
• Proxy configuration (proxyConfiguration) — Apify Proxy, defaults to RESIDENTIAL. Used for the YouTube path; Reddit's public API needs no proxy.

Output fields

One row per comment: platform (youtube/reddit), content_url (the source video/post), comment_id, comment_url (direct link to the comment), author, author_url, text, likes (YouTube like count / Reddit score), reply_count, posted_at, sentiment, flag, and scraped_at. sentiment/flag are populated only when AI tagging is switched on — otherwise they come back empty.

Pricing

Pay-per-result — you're charged one comment-result event per comment returned (unresolved/skipped items are not charged). See the Store pricing tab for the current per-comment rate.

Notes

• Reddit posts are fetched entirely via a public API — no login or proxy required for that side.
• YouTube goes through Apify Proxy (RESIDENTIAL by default).
• The AI add-on never blocks output: if no API key is present, comments still ship with a deterministic sentiment/flag rather than an error.
• TikTok, Instagram, and Facebook comments are not covered by this actor — only YouTube and Reddit.