免费查找任何视频片段的原始来源。
The Video Source Finder is a free online tool that lets you instantly detect the original video behind any clip, short, or repost. Whether you stumbled across a viral snippet on TikTok, a cropped clip on Instagram Reels, or a YouTube Short with no credit, simply paste the link and our tool will analyze the video
The tool uses yt-dlp to fetch the video's metadata from the URL you provide. Platforms like YouTube embed source information in clip metadata — for example, a YouTube Clip always points back to the full video URL. The tool extracts this information and surfaces the original video link instantly. No video download occurs.
The tool supports YouTube (including Clips and Shorts), TikTok, Instagram Reels, Twitter/X, Vimeo, Facebook, Twitch, Dailymotion, and dozens of other platforms supported by yt-dlp. If a platform has a publicly accessible video URL, it will likely work.
Not every platform embeds source information in its metadata. TikTok reposts, Instagram shares, and many other clips do not carry a direct reference to their origin. In those cases, the initial free scan can't detect a source — that's when the AI Deep Research feature becomes valuable.
Deep Research downloads the first 90 seconds of the video's audio and transcribes it using OpenAI Whisper. It then sends the transcript to GPT-4o-mini, which extracts 2-3 highly distinctive search queries based on specific names, quotes, or events mentioned. Those queries are run against YouTube search, and GPT ranks the results to identify the best match, returning it with a confidence percentage and explanation.
Yes, the basic metadata-based source detection is 100% free with no account required. The AI Deep Research feature costs 1 credit, which you can obtain through the credits system. New users receive 15 free credits upon signing up.
Direct metadata detection only works when the platform stores a reference to the original. For re-uploads with no metadata trail, use the AI Deep Research feature — it transcribes the content itself rather than relying on metadata, making it effective even for videos that have been reuploaded or cropped.
Accuracy depends on the content of the video. Videos with unique spoken phrases, named people, or specific events are matched with high confidence (often 80–95%). Videos with generic content or no speech will have lower confidence. The confidence score and AI explanation help you judge whether the match is reliable before clicking through.
The tool relies on publicly accessible metadata and audio. Private, members-only, or age-restricted videos may not be fully accessible without authentication, in which case the tool will return an error. It works best with public video content.
Google Lens, TinEye, and Yandex Images search for visual matches to the video's thumbnail image — useful if the original video has a distinctive or unique thumbnail. Deep Research, on the other hand, analyzes the audio content of the video itself, making it more effective for clips where the visual thumbnail has been cropped, changed, or is generic.
The tool is designed to trace from a clip outward to the source — meaning you'd provide the suspected copy's URL. If the copy is on a platform that embeds source metadata, it will quickly link back to your original. For deeper detection across platforms, combine this tool with reverse image search and manual YouTube search using your video's unique phrases.