Use when a video is edited and the title, thumbnail text and description are still undecided.
youtube-upload-packaging.md
You are packaging a finished video for YouTube. You are not editing the video and you are not rewriting the script.
Transcript or detailed outline: {{VIDEO_CONTENT}}
Who searches for or gets recommended this: {{AUDIENCE}}
Our last five titles and how each performed: {{PAST_TITLES}}
Link we want clicked, if any: {{TARGET_LINK}}
Output five numbered sections.
1. Ten title options, each under 60 characters, each promising something {{VIDEO_CONTENT}} actually delivers. Label each search-led or browse-led.
2. A table: Title | Thumbnail text (max 4 words) | What the pairing promises | Why it would not fit another video. Thumbnail text must not repeat words from its own title.
3. The description. The first two lines work as the collapsed preview and do not open with the channel name, then a paragraph on what the video covers, then {{TARGET_LINK}} with one line saying what is behind it.
4. Chapters as timestamp plus label, taken only from what {{VIDEO_CONTENT}} shows happening. If the source carries no timestamps, say so and give the labels in order without times.
5. A pinned comment asking {{AUDIENCE}} something they can answer from their own work.
Check every option against {{PAST_TITLES}} and reject any pattern that already underperformed there. No promise the video does not pay off, no all caps, no em dashes.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When should I use this rather than asking for title ideas?
When the edit is locked and title, thumbnail text and description are all still open. Ten title ideas on their own give you no way to choose between them. This pairs each title with thumbnail text, states what the pairing promises, and checks options against titles that already underperformed on your channel.
What do I need in front of me?
The transcript with timestamps, or section four comes back without times and is unusable. Your last five titles with click-through rate beside each, since the pattern check is guesswork without the numbers. Also who finds this video, and the link you want clicked with something behind it worth describing.
What comes back, and which part do I work from?
Ten titles labelled search-led or browse-led, a pairing table, the description, chapters and a pinned comment. Work from the pairing table. A title and a thumbnail chosen separately end up making two different promises about one video, and the viewer resolves that by not clicking either.
What is the mistake?
Picking the title that promises most. The prompt rejects promises the video does not pay off, and an overselling title lifts click-through then loses the retention that decides whether the video keeps being shown. Check the chapter labels against the video as well, since they come from the transcript alone.