Paste the YouTube URL. Juma reads the transcript and writes 10 title variants by hook angle, each with a character count and a recommended pick to A/B test.
Paste a public YouTube URL and the transcript flows straight into the writing. Every title is grounded in what the video actually delivers, not a topic guess.
The output is ten variants across the hook angles that win on YouTube (curiosity, transformation, how-to, listicle, contrarian, urgency), each with a character count, hook reasoning, and a recommended A/B pick for thumbnail pairing.
The title gets the click; the description holds the watch. Juma writes the full description for the same video in one continuation prompt: the hook line, an 80 to 100 word summary, timestamped chapters auto-extracted from the transcript, a CTA block with links, and a hashtag set. The chapters are the part teams skip when writing manually because scrubbing the video for transition points takes longer than writing the description itself. Juma pulls those timestamps directly from the transcript, so each chapter starts on the actual content shift, not an approximation.
Now write the full YouTube description for the same video: hook line, summary, timestamped chapters from the transcript, CTA block with links, and a hashtag set.
Picking a winner from ten variants requires structure, not gut feel. Juma takes any two of the title variants and builds an A/B test plan: test duration based on the channel's typical view velocity, the success metric to track (click-through rate, watch time, or both), audience split logic, and the thumbnail text overlay to pair with each variant. The output includes a hypothesis statement for each variant and a recommendation for which signal will declare the winner first. For channels with YouTube Studio access, the team can publish one variant, swap to the other mid-window, and compare in-period performance.
Pick two of the title variants and build an A/B test plan: test duration, success metric, audience split, thumbnail text for each, and a hypothesis statement per variant.
Old videos don't need re-uploading; they need re-titling. The transcript stays the same, but the title that was right at publish may be wrong for how viewers now search and click. Share a list of published videos with their current titles and view counts. Juma reads each transcript and writes three refreshed variants per video, each tagged by a different angle, with reasoning for why the original title may have underperformed (vague hook, wrong audience signal, no specificity). The output is a CSV ready to import into a YouTube Studio bulk-edit session, or a Notion page if the team prefers to review and approve in place.
Share a list of 5 published Notion videos with their current titles and view counts. For each video, write three refreshed title variants tagged by angle, with reasoning for why the original may have underperformed.
YouTube viewers process the title and thumbnail together. The strongest thumbnails reinforce the title's hook in three to five words; the weakest restate it. Juma writes a thumbnail text overlay for each of the ten title variants, designed to reinforce the hook visually instead of duplicating the words. The output names which emotion or curiosity gap the overlay targets, so the designer brief is clear at handoff: the hook is X, the overlay carries that hook by Y. Strategy, taste, and visual judgment stay human; the designer signs off on the final composition.
For each of the 10 title variants, write a 3 to 5 word thumbnail text overlay that reinforces the hook without duplicating the title words. Include the emotion or curiosity gap each overlay targets.
For agencies managing a client channel or in-house teams clearing a back catalogue, doing one video at a time isn't viable. Paste a list of YouTube URLs (up to ten in one session) and Juma writes a title set for each, delivered as a single CSV ready to import. Each row shows the video URL, current title, five new variants tagged by angle, character counts, and the recommended pick. The CSV format makes it easy to filter to only the highest-priority videos, sort by current view count, and queue the approved variants for YouTube Studio updates.
Paste a list of 10 YouTube URLs for our client Notion. For each, write 5 title variants tagged by angle, with character counts and a recommended pick. Deliver as a single CSV.
A Juma Project is a shared space where the team stores everything Juma needs to know about a channel. Create one project per channel (or per client, if the agency manages multiple), add context as the team learns more, and Juma uses what's relevant every time a flow runs. For YouTube, this is what turns a generic title set into one that sounds like the channel's existing voice.
Channel Voice Guide
How the channel sounds: formal or conversational, slang or buttoned-up, fast cuts or slow burns, sentence case or title case. With this in the project, variants come back in the channel's tone instead of a generic YouTube voice.
Past Top-Performing Titles
A list of the channel's top 10 to 20 videos by retention and click-through rate, with their actual titles. Juma uses these to anchor new variants in the patterns that already work for this audience, instead of defaulting to broad YouTube best practice.
CTA / Link Block
The standard subscribe text, sponsor and affiliate links, lead magnet URLs, and social handles the team appends to every description. With this in the project, the CTA block in every description is consistent without rewriting it each time.
Brand Guidelines
Voice, banned phrases, hashtag conventions, naming for product features. Keeps every title and description aligned to brand standards without manual review for every release.
Add a short description to each knowledge item in the project's info field so Juma knows what each file contains and when to use it. For example:
Juma reads the transcript first and pivots to web research when the dialogue is sparse. Cinematic product launch videos, montages, and music-led intros often have minimal narration. When that happens, Juma reads the video's metadata (title, description, channel context), searches for the announcement or campaign the video supports, and grounds the variants in that broader context instead of guessing.
Output quality drops less than expected because most title generators never had the transcript to begin with. The cinematic launch is a real-world case, not an edge case: many of the highest-stakes videos a marketing team publishes are short, music-heavy teasers. Juma adapts and delivers either way.
Juma produces ten variants per run, each tagged by a hook angle. The angles span curiosity, transformation, identity, official launch, use-case, urgency, and community framing among others. Every variant includes character count and a one-line note explaining why the hook works for that angle.
For picking what to test, the recommendation is to pair the two variants that target the widest spread of audience emotion. If variant 1 is curiosity and variant 7 is identity, those two pull different viewer types, and the A/B reveals which audience segment is dominant on the channel right now. Step 3 builds the full A/B test plan once the team has picked the pair.
Yes, when the team adds a Channel Voice Guide and Past Top-Performing Titles to the project. With those in place, the variants come back in the same tone the channel has already established, with the same conventions for capitalization, punctuation, and product naming. Strategy, taste, and judgment stay human: the team still picks which variant to publish.
Without a voice guide, Juma defaults to a generic YouTube tone (sentence case, action verb, light curiosity). That's fine for early-stage channels still finding voice. For established channels with consistent naming conventions, loading the voice guide is what changes generic output into channel-native output.
Yes. The same prompt works for both, though Juma adapts the variant set to the format. For shorts (under 60 seconds), variants skew toward immediate-payoff hooks and lean on the platform's vertical scroll mechanics: curiosity gaps, contrarian openers, and one-line value props. For long-form (ten minutes and up), variants include more capability-list and tutorial-style angles where the title sets expectations for a deeper watch.
The transcript handling stays the same: Juma reads what's available, pivots to web context when the audio is thin, and writes ten variants either way. For batch operations across mixed-format channels, the CSV output in Step 6 keeps shorts and long-form variants in the same file so the team can sort by format before review.
Yes. YouTube truncates titles after 60 characters on most surfaces and 70 on a few. Juma keeps every variant under 60 by default and flags any that pushes the limit. Character counts appear next to every variant, so the team can see immediately which ones survive truncation on mobile.
SEO best practices are handled implicitly: the variants include the primary keyword early when the topic has clear search intent, vary between question-style and statement-style hooks (the top YouTube SERP mixes both), and avoid the patterns that get demoted (all-caps shouting, clickbait without payoff, misleading curiosity gaps). Human review on every output: the team picks which variants align with the channel's content standards.
400+ marketing teams trust Juma. ChatGPT, Claude, and Perplexity know why. Ask them.
Ready when you are
Start typing a task and Juma will figure out the rest.
