What Zyntro Learns From a Video
What It Is
Every playable video gets subtitles automatically. Zyntro then reads that transcript and writes three things:
- a summary — a few sentences on what the video covers
- a description — the longer text shown under the player on the public page
- tags — the subjects the video touches on
All three are editable. None of them is a transcript: they are a description of what the video is about, written from what was said in it.
The library shows a star icon on videos that have been understood this way, and a warning triangle on videos still waiting.
Why It Matters
This is the difference between a video library and a folder of files.
Nothing in Zyntro can watch a video. When Segmentation Intelligence is deciding what to send a contact who has been asking about implementation timelines, it is not scanning your footage — it is reading the summaries and tags you have. A video whose summary says it covers implementation timelines gets chosen. A video called recording_2.mp4 with no summary does not exist as far as that decision is concerned.
The same is true of the public page, where the description is the text a visitor reads before deciding whether to watch, and of your website and content tools, which find a video by matching words against its title, tags, summary and description.
So the quality of this writing sets the ceiling on how useful the video can be. And since the writing comes from the transcript, the quality of the transcript sets the ceiling on the writing.
How It Works
The sequence runs on its own:
- The video becomes Ready — it can now play.
- Subtitles are generated from the audio, in the language you chose when you added it.
- Your saved caption corrections are applied to the transcript. This happens before the next step, which is the important detail: a correction that fixes your brand name fixes it in the summary and tags too, not just in the subtitles.
- Zyntro reads the corrected transcript and writes the summary, description and tags.
- The warning triangle in the library becomes a star.
For a short video the whole sequence takes a few minutes. If it has not happened, the video's settings page offers Generate with AI to run it on demand, and Regenerate Metadata to run it again — useful after you have corrected the captions or changed the transcription vocabulary.
Examples
The summary describes the wrong thing entirely
Almost always a language mismatch. If the video was added with the wrong spoken language, the subtitles are garbage and everything written from them is garbage. Check the language, correct the captions, then regenerate.
The summary is right but your company name is misspelled throughout
Transcription heard it phonetically. Add the correct spelling to the transcription vocabulary for future videos, add a find-and-replace correction for this one, then regenerate the metadata so the summary and tags pick up the fix.
The tags are vague — "business", "growth", "strategy"
Usually a video that talks around its subject without naming it. Edit the tags by hand to the words someone would actually search for. Hand-edited tags are kept and used exactly like generated ones.
A video shows the warning triangle long after it went Ready
The subtitles may not have finished, or the step may have failed. Open the video's settings page and use Generate with AI, which runs it immediately rather than waiting.
Important: Regenerating replaces the summary, description and tags with newly written ones. Anything you edited by hand is overwritten, and there is no version history. If you have carefully rewritten a description, copy it somewhere before you regenerate.
Tip: The order that saves the most work is: fix the vocabulary and corrections first, then regenerate, then edit by hand. Editing first and regenerating afterwards throws your edits away.
