Title Generator

What happens after you press Generate

Your topic and the chosen type are posted to this site's /api/ai endpoint. The server wraps them in a fixed instruction (“Write 8 varied, catchy blog post titles about this topic”) and forwards that to a hosted language model through Groq's API, so the topic does leave your device and is processed by a third-party AI provider. The site keeps no record of it, but it isn't a place for confidential project names.

The model is asked to answer in the same language as the topic and to return only a numbered list. The page strips the numbers and any surrounding quotation marks before showing the rows. Generation runs at a high randomness setting (temperature 0.9), which is why two runs on the same topic rarely overlap.

What each title type asks for

Limits you won't see in the box

• The topic is cut to 1,000 characters on the server. A counter under the box shows how much of that you have used and turns red once the brief is longer than what will be sent, so keep it to a few lines.
• The reply is capped at 500 tokens, which is roughly 350 words. Eight titles of 10–15 words use about a third of that, so the cap only bites if the model pads the list.
• The site allows 12 AI requests per minute from one IP address. Past that you'll see “Too many requests. Wait a minute and try again.”
• When the shared free AI quota is exhausted the message changes to “We've hit the free AI limit for now.” Wait a little and retry; nothing is wrong with your input.
• Occasionally the model returns an intro line or a heading before the list. The parser keeps every non-empty line, so a stray line can appear as a “title” — ignore it.

Writing a brief that produces usable titles

Compare two briefs on the same subject. “running tips” gives the model nothing to hold on to, so the eight titles come back generic. “Beginner running plan for people over 40 who hate the gym, keyword: couch to 5k” fixes the audience, the objection and the search phrase, and the batch reflects all three.

The generator does not count characters or check search rankings. Google trims most result titles at around 600 pixels wide, which is usually 50–60 characters, so run your favourite through the Character Counter before you commit to it. If a title is close, shorten it yourself rather than regenerating — the model has no memory of the previous batch.

For a listicle, put the number in the brief (“7 mistakes”); for a how-to, say so. Mixing the first half of one title with the second half of another is normal practice and often beats any single suggestion.

When the button does nothing or the list looks wrong

• The button stays disabled until the topic box contains something other than spaces.
• “AI tools aren't enabled yet” means the site's AI key isn't configured on the server, not that your request was bad.
• “The AI request failed. Please try again.” is an upstream error from the model provider; a second attempt usually succeeds because the server tries a chain of models.
• Titles in the wrong language: the model follows the language of the topic, so write the brief in the language you want the titles in.
• Fewer than eight titles: the model sometimes merges or skips entries. Run it again; each batch is independent.