Titles that keep the promise the thumbnail made.
Generate from a topic, or hand it the thumbnail and let a vision model write titles that complement the image instead of repeating it. Then take any title straight back into a thumbnail request.
Twelve strategies
Pick the angles, not just the count.
Curiosity, authority, breaking news, educational, contrarian, list, challenge, story, comparison, urgency, emotional, minimalist. Turn off the ones that are wrong for your channel and the batch spreads across what is left.
Every title is judged
Character count, and whether it survives mobile truncation: fits on mobile, truncates on mobile, or truncates almost everywhere.
Eight variations
Shorter, more curious, more professional, more dramatic, less clickbait, stronger keyword focus, rewrite, generate similar. One cheap call each.
Reasons, not just output
Each title comes back with its strategy and one line on why it works, so you are choosing rather than guessing.
Free on your own machine
Titles are the cheapest thing here and need no image provider at all. Point them at Ollama and the whole feature costs nothing per run.
Both directions
Title to thumbnail, thumbnail to title.
A title you like becomes an idea and a headline on the Create screen with one click. A thumbnail you already made becomes the input a vision model reads before writing titles for it. They are the same project, not two tools that happen to sit next to each other.