Sources
YouTube retrieval uses the official YouTube Data API. TikTok retrieval uses a third-party data provider because the ordinary commercial TikTok Display API does not expose arbitrary public comment bodies.
Normalization
Provider-specific responses become one schema for IDs, authors, text, dates, likes, reply counts, parent relationships and stable links when available.
Coverage
The workspace shows loaded comments, remaining allowance, available pagination and partial-result warnings. Reply threads are loaded separately. Search fields filter loaded top-level comments; reply text is read inside its thread or searched in a downloaded file. A source comment count is not a guarantee that the same number of rows can be retrieved.
Deduplication and exports
Rows are deduplicated by platform, source and comment ID. The free basic CSV contains up to 100 rows from the current filtered view, including replies already opened in that view. Advanced Exporter jobs are a separate paid workflow for larger targets, reply traversal and field choices. Reusing a recent saved dataset does not call a provider again.
What zero means
An empty result is only treated as confirmed when the upstream provides a reliable reason. Ambiguous empty responses are surfaced as partial or unavailable, not as proof that the video has no comments.