One writer. You guess.
Five titles from one model, in one voice. No second opinion. You choose on a hunch.
For YouTube creators / free
Don't settle for
one AI's
YouTube titles.
Your title and thumbnail decide who clicks. Let the best frontier models compete for them. Paste your transcript. Up to 25 models write the whole package. A blind judge picks the winners. You make the call.
Free to use. You bring your own OpenRouter key and pay OpenRouter directly.
The problem
One chat window.
One voice.
You finish the edit. Then the second job starts. A title. A thumbnail. A description, tags and an email to your list.
So you paste the transcript into one chat and ask for five titles. They come back fine. They also sound alike. One model wrote all five.
You pick the least bad one and upload. Days of work go out under a title you never loved.
Your video earned better than the first draft of one model.
Before / after
Pick from the best, not the first
Frontier models write differently. Put them side by side and the strong ideas stand out. A judge that can't see the authors keeps it fair.
Fourteen writers. A fair judge.
Fourteen models took the job. Thirteen finished, with 65 titles. JEV scored every one without knowing who wrote it. You get the top three, and the other 62 in rank order.
The bridgePackage Desk runs the panel and the judge for you. One transcript in. The best package out.
Sign up freeProof / November 12, 2026
A real run, replayed
This run was recorded on November 12, 2026. Fourteen models wrote from one transcript. JEV judged what came back. The times, costs, picks and percentages are real. One model ran out of time. It stays in. The transcript is a made-up script about this tool.
Fourteen writers get the same transcript at the same time.
Transcript540 words
Writers
In the pool: 65 titles, 39 prompts, 13 descriptions, 13 tag sets, and 13 emails.
65titles from 13 models (one more timed out)
3picked by JEV, blind
$0.0071what the writer of winning title #3 cost in this run
$0.13for the whole run, judge included
Why it works / five steps
Many writers. One fair judge.
Every writer gets the same prompt at the same time. The judge never sees who wrote what. You make the final call.
Bring any transcript
Paste it, or drop in a .txt, .srt or .vtt file. Timestamps and cue numbers come out for you.
Hear many voices
Up to 25 models get the same prompt and transcript at once. Each writes a full package. None sees the others' work.
Lose the repeats
Candidates are pooled by section. Word-for-word duplicates merge, and every model that wrote one gets the credit.
Get a fair judge
JEV sees each section shuffled, with the names removed. It scores every candidate. It never knows which model wrote what.
Keep the choice
You get the top three titles, the top two thumbnail prompts, and the best description, tag set and email. The rest stay one click away, in JEV's order.
If JEV can't be reached, nothing is picked. The page tells you why. No other model stands in.
Pricing / you pay providers directly
What it costs
Package Desk is free. You pay the model providers directly, at their prices. Your cost depends on the models you pick and the length of your transcript. Each writer's cost shows as it finishes.
Free
Bring your own OpenRouter key. You pay this site nothing.
$0.13
What one run cost on November 12, 2026, with 13 of the 14 writers finishing.
$0.042per million input tokens
Per TypeSafe's docs. In the run above that came to $0.0008.
Privacy / bring your own keys
Your keys stay with you
One OpenRouter key runs it all: the writers and JEV. Add a TypeSafe key if you have one, and JEV runs at TypeSafe directly. Your keys stay in your browser. So do your past runs.
When you start a run, your keys and transcript pass through this site's server to OpenRouter, and to TypeSafe if you added that key. The server stores nothing.
Your next upload
Give your next video
its best title.
Have a transcript and your OpenRouter key ready.