Is this AI tweet generator free?

Yes. Three generations per day, each returning three scored post variants, with no account, no email address and no credit card. The daily cap exists because every generation costs real model tokens, not because there is a hidden upgrade step inside the tool.

Do I have to sign up or connect my X account?

No. The generator never touches your X account and never asks for a login. You type a topic, you get three drafts, you copy the one you like and post it yourself.

What is the anti-slop gate?

Every draft is scored 0 to 100 by the same Distribution Score engine VoicePost runs on paid drafts. Anything below 70 is sent back to the model once, with the exact failed checks attached, and only the better version survives. Most free generators show you the model’s first attempt with no check at all.

Why is 70 the passing score?

The scorer starts every draft at 100 and subtracts 25 points per hard failure (em-dashes, hashtag spam, banned AI phrases, failed burstiness) and 10 per soft failure. Below 70 a post has broken at least one hard rule or several soft ones, which in practice is the band where posts read as machine-written.

Can I use this as a Twitter thread generator?

Not directly. It writes single posts, because a thread that opens with a weak first post is a thread nobody reads. Generate the strongest opening post here, then write the follow-ups yourself while the angle is fresh.

Will AI-written posts hurt my reach on X?

Generic ones will. X ranks on engagement signals, and posts carrying the standard AI tells (hashtag stacking, throat-clearing openers, the "not X, it is Y" construction) get scrolled past, which the ranking system reads as low quality. That is exactly what the score in front of each draft is measuring.

Can X or an AI detector tell that a post was generated?

X does not run a public AI detector, and third-party detectors are unreliable on text this short. Readers are the real detector: they recognise the patterns without naming them. Edit any draft until it sounds like something you would actually say out loud.

Are the generated posts mine to use?

Yes. Copy them, edit them, post them, use them commercially. Nothing you type is used to train anything, and nothing is stored against an account, because there is no account.

Free AI Tweet Generator

An AI tweet generator turns a topic into a ready-to-post draft for X. This one is the only free generator that scores its own output before showing it to you: every draft runs through an anti-slop scorer, and anything under 70 out of 100 gets rewritten once before it reaches the page. You see the score on each variant, so you are never guessing whether a post reads as machine-written. No signup, no login, three free generations a day. VoicePost (voicepo.st) is the product this scorer comes from: it drafts posts in your own trained voice every day, and you approve each one before it publishes now or on the schedule you set.

What is the post about?A topic, a lesson, a thing that happened today. One or two sentences is plenty.

ToneLength

Free, no signup.

How it works

You give the generator three things: a topic, a tone, and a length. It writes three different posts about that topic, scores each one, sends the failures back for a single rewrite, and hands you what survived with the numbers visible. Two model calls maximum per generation, which is why the daily cap is three and not thirty.

What the scorer is actually checking

The score is the Distribution Score, a pure-code scorer that runs thirty named checks against a draft. It starts at 100 and subtracts. Hard failures cost 25 points each and are the things that mark a post as machine output at a glance:

Em-dashes. The single most reliable AI tell in short-form text. Human posts use commas, full stops and line breaks.
Hashtags and emoji stacking. One emoji is a person. Four is a content calendar.
Banned phrases. A kill list of the words that only appear in generated text: delve, leverage, seamless, game-changing, groundbreaking, cutting-edge, robust, scalable, paradigm shift, and about a dozen more.
Failed burstiness. Humans write sentences of wildly uneven length. Models write sentences of almost identical length. The scorer measures the standard deviation of sentence length and fails a draft whose rhythm is too flat.

Soft failures cost 10 points each and cover the subtler patterns: throat-clearing openers, the “not X, it is Y” contrast, rhetorical setups like “here is the thing”, adverb intensifiers, three-item lists where two would do, and abstract nouns doing work a concrete detail should be doing. You can run any text through the scorer on its own with the AI slop checker.

Why the gate is set at 70

70 is the pass threshold the paid engine uses. Below 70 a draft has either broken one hard rule outright or accumulated several soft ones, and in the corpus the scorer was calibrated against, that band is where posts stop sounding like a person. Between 55 and 69 a draft is usually salvageable with one edit. Above 70 it reads clean. Every score you see on this page is the real number, including when a draft came back from the rewrite pass still imperfect. Hiding a bad score would defeat the point of showing one.

The three tones, and when each one works

Direct states the useful thing immediately, with no windup. It suits how-to observations, technical notes, and anything where the value is the information itself. It is the safest default and the hardest to make boring, because there is nowhere to hide.

Story writes the post as something that happened to you, in first person, with one concrete detail. It suits lessons, mistakes, and build-in-public material. Stories travel further than advice, because a stranger can picture the events even if they disagree with the conclusion. It is also the tone that needs the most editing, since the generator does not know your actual week.

Contrarian leads with a position and argues it. It suits opinions your work has earned. Use it carefully: a take you cannot defend in the replies costs more than the reach it buys.

How to edit a draft up to a post you would actually send

Treat the output as raw material. Three edits fix almost everything a generic generator produces:

Replace the generic noun with your real one. “a feature” becomes the feature. “a customer” becomes the customer who emailed you on Tuesday.
Put a real number in. Not a rounded one. 47 minutes reads true, 45 minutes reads invented, and the scorer penalises suspiciously round figures for exactly that reason.
Cut the last sentence. Models close by summarising what they just said. Human posts end on the sharpest line and stop.

What a free generator cannot do

It does not know your voice, your niche, or what happened in your industry this morning, so every draft here is a competent generic post about a topic rather than a post only you could have written. That gap is the whole reason the paid product exists, and it is worth being honest about rather than pretending a topic box can close it. If you want the full picture on generating material when the box is blank, read what to post on X when you have nothing to say.

Frequently asked questions

Is this AI tweet generator free?: Yes. Three generations per day, each returning three scored post variants, with no account, no email address and no credit card. The daily cap exists because every generation costs real model tokens, not because there is a hidden upgrade step inside the tool.
Do I have to sign up or connect my X account?: No. The generator never touches your X account and never asks for a login. You type a topic, you get three drafts, you copy the one you like and post it yourself.
What is the anti-slop gate?: Every draft is scored 0 to 100 by the same Distribution Score engine VoicePost runs on paid drafts. Anything below 70 is sent back to the model once, with the exact failed checks attached, and only the better version survives. Most free generators show you the model’s first attempt with no check at all.
Why is 70 the passing score?: The scorer starts every draft at 100 and subtracts 25 points per hard failure (em-dashes, hashtag spam, banned AI phrases, failed burstiness) and 10 per soft failure. Below 70 a post has broken at least one hard rule or several soft ones, which in practice is the band where posts read as machine-written.
Can I use this as a Twitter thread generator?: Not directly. It writes single posts, because a thread that opens with a weak first post is a thread nobody reads. Generate the strongest opening post here, then write the follow-ups yourself while the angle is fresh.
Will AI-written posts hurt my reach on X?: Generic ones will. X ranks on engagement signals, and posts carrying the standard AI tells (hashtag stacking, throat-clearing openers, the "not X, it is Y" construction) get scrolled past, which the ranking system reads as low quality. That is exactly what the score in front of each draft is measuring.
Can X or an AI detector tell that a post was generated?: X does not run a public AI detector, and third-party detectors are unreliable on text this short. Readers are the real detector: they recognise the patterns without naming them. Edit any draft until it sounds like something you would actually say out loud.
Are the generated posts mine to use?: Yes. Copy them, edit them, post them, use them commercially. Nothing you type is used to train anything, and nothing is stored against an account, because there is no account.

By hand vs with VoicePost

You can do all of this yourself. Here is what it takes to keep it up every day, and what VoicePost handles for you.

	By hand	With VoicePost
Coming up with the post	Staring at the compose box, then posting something generic or nothing at all.	Drafts arrive every day, built from what actually happened in your niche that morning.
Sounding like yourself	Editing a generic draft by hand until the phrasing stops feeling borrowed.	The draft is written against a voice profile trained on your own posts, so most of that editing never has to happen.
Catching AI tells	Re-reading each draft and hoping you notice the em-dash and the flat rhythm.	The same scorer runs on every draft before you see it, and low scorers are rewritten automatically.
Posting consistently	Remembering to write, then remembering to post, on the days you have time.	You approve a draft once and it publishes now or on the schedule you set.

Read next

What to Post on X When You Have Nothing to Say

Five formats that generate themselves from your workday, plus the capture habit that ends the blank compose box for good.

Last reviewed: July 2026