AI Carousel Generator for Instagram & TikTok

Overall: very simple, very fast to get started, and that's your biggest strength.

1. Would you actually post the result under your own account?
I tested it on my own real problem, which isn't one of your suggestions: I don't know what carousels to make that actually get engagement. You'll see the exact prompt in my screenshot.

Result: no, I wouldn't post it. For my use case it failed, because it doesn't know my brand:

• Style: my brand is light, the carousel came out fully dark
• Voice: it couldn't access my Instagram, so it didn't sound like me
• Storyline: "Launch your app → Nobody showed up → Ask your friends → Bugs. Strangers find them first." The first two slides make you swipe, but then it's problem after problem and never clearly says we are the solution
• Mascot: didn't fit my brand

2. Does it look like AI slop anywhere?
No. The blog post test (gym article) looked real, people even looked like they were in the same gym. Mine wasn't slop, just off-brand.

3. Did the remake feel close to the reference style?
Didn't test a carousel remake. I used a blog post link and my own prompt (see screenshot). The blog post one worked well.

4. Did the single-slide edit work?
Yes, it changed only that slide and did what I asked. 👍

5. Where did you get stuck, wait too long or feel confused?

• Sign-up comes too early. I typed a prompt on the landing page and had to log in right away. People want to see a result first. Show a (partial/blurred) result, then say "unlock this", like image-generation tools do. This is where you'll lose most people
• Voice input works on the landing page but not in the app
• Bad input = bad output. Most people won't read "Describe a post or paste a link". Help them give good input

6. Anything that broke?
No: links loaded, no cut-off text, exports and zip worked. On the landing page, Attach, Link and size are clickable but do nothing.

The #1 thing I'd add: a feedback module 👍👎
When the carousel is ready, ask: thumbs up or thumbs down? On thumbs down, ask what they don't like and show the components: color palette, text, storyline, images… Then ask what they actually wanted. After that:

• Tell them what's missing on their side: "You haven't added your brand settings yet", "Add the link to your website", "Connect your Instagram". Most people put bad info into AI and get a bad result back. This shows them how to fix it
• Actually generate a better version based on their answer: "This is how we'll make it good", then make it
• Save the feedback either way, even if you don't use it yet. Do it per carousel for now. Later, if everyone says the same thing, you know exactly what to fix for everyone. It can't hurt you

This is also the perfect moment to ask for payment: they've just seen it get better.

This is what turns Tweenery from a tool a few experts use into something lots of people use. Don't even hesitate, add it.

On-brand + engaging is the real value

• Brand setup: just ask for my website and pull colors, fonts and tone from it. For voice, connect Instagram (e.g. via Windsor.ai)
• Color palette: let me adjust it
• No onboarding questions up front. Let people fix things after seeing a result, and ask for payment then: "All your slides can look this good, just upgrade"
• Engagement: my real problem is not knowing which carousels to make. Everyone says carousels drive engagement, but I haven't had success. My Claude Design carousels looked great but got zero engagement, and I had to think out every word myself. If you give me the ideas and bake in what actually converts, that's what I'd pay for. And tell people "we did it this way because this is what works". No promises, but it shows you're the expert

Landing page

• Answer "Why not just use ChatGPT?" It's clear what you do, not what problem you solve. Mine: I don't want to spend time on ideas, I want content that actually gets engagement
• Explain it like I'm 5: "We make your Instagram carousels for you". Some easy SEO wins too
• Light mode: only showed dark for me, maybe follow system settings
• Pricing: $19 / $49 makes sense, but let free users set up their brand kit. Once my brand is in, I'm hooked
• Main CTA: the single ChatGPT-style input box and the suggestions are very good. Keep Start free / Log in as is

Small stuff

• Model picker: nice, explain which model is best for what
• Aspect ratios with "Instagram feed / TikTok" labels: exactly what I need
• Loading: remove the blue bar sliding over the text box. The most basic car has the best engine
• Add a share option for chats, happy to share mine

Most of it works and it's very simple. I think a lot of people will sign up. For me it just needs to be on brand and engaging. Happy to go deeper on anything!

Q: Would you actually post the result under your own account? If not, what stops you?
A: Oh yes, I would. My brand is clean, minimalistic and it was able to create carousels similar to that. A couple of clinical claims on the carousels that I did verify but other than that, yes, I would be happy to use and post under my own account

Q: Does it look like AI slop anywhere? Point to the slide.
A: No, does not look like AI slop, it is clean and modern.

Q: Did the remake feel close to the reference style?
A: I did not request a remake. I started from scratch so unsure how that would have worked.

Q: When you asked for a single-slide edit, did it change only that slide and do what you meant?
A: I did not ask for an edit because to me it was almost 90% done. Only change I could have made was to change the fonts but thats almost going to work 100% so did not try.

Q: Where did you get stuck, wait too long or feel confused?
A: I did not get stuck, did wait for the results for around 3 mins but it was worth the wait

Q: Anything that broke: links that wouldn't load, cut-off text, failed exports.
A: Nothing broke, process was smooth and sweet.

actually looked great. but it is ai. so i wonder how is it different from ai slop you want to replace. also an idea of token consumption would be nice before proceeding

First impressions:
The hero is strong - "Your AI carousel designer. It does all the work.", with the prompt box right in the hero. "No credit card" is visible.
On first load the hero was blank dark blue for a few seconds, with only the nav showing. Scrolling into "Every slide, designed for the swipe" showed empty slide tiles, then a fully black screen for a moment. It looked broken. Lazy-load the images or show placeholders.
Pricing shows yearly-upfront amounts by default, and the page never says how many credits one carousel costs.

Signup:
I typed a prompt in the hero and hit Enter. It took me to signup with my prompt kept in the URL. Nice.
After Google signup and the "Set up your brand first?" screen (good choice: Set up my brand vs Skip for now), my prompt was gone. The box was empty. Carry it through.

What I made:
"A 5-slide carousel on 3 study habits for high school students". You can watch the steps live (plan → find a look → build slides → check), and it took 1:13.
The result was genuinely good: a "3 Study Habits Cheat Sheet" cover, a clean paper look with a lime highlight, and useful copy. The agent explained its assumptions (Instagram, placeholder @yourhandle) and suggested next steps.
Slide 3 had a big empty gap between the subtitle and the body text.
Edit: "Make the headline on slide 3 bigger". It took 21s, changed only slide 3, and the agent warned me that "cram" now sat alone on the second line. That honesty is great.

Credits:
The first carousel cost 52 of my 300 free credits and the edit cost 7. So the free plan is about 5 carousels. Say that on the pricing page.

Bottom line:
This is the best result I've had from an AI carousel tool. Fix the first-load blank screens, keep the prompt through signup, and show the cost per carousel.

Tweenery — honest first impressions

Hi,

I spent a good session with Tweenery and followed your test instructions. Short version: the core idea works, and the output is better than I expected. A few things hold it back from being something I'd trust every day.

What worked well

• Signup with Google was quick, no card, 300 free credits. Good start.
• Onboarding is optional, which I liked. My choices (goal, audience, voice) actually showed up later in My setup, so it feels like it's listening.
• The gym topic gave me a six-slide 4:5 carousel in a calm cream/terracotta editorial style with an illustrated cover. The story flowed well, from the hook to practical steps to a save-for-later recap. No cut-off text, no gibberish. Would I post it? For a general account, yes, after a quick fact check. It doesn't look like AI slop, but it also doesn't look like my brand yet.
• The single-slide edit is the best part. I asked to make slide 2's hook punchier, got "Go all in, burn out fast." in about 40 seconds, and the other slides stayed untouched. Exactly what I meant.
• The first ZIP export worked: six clean PNGs.

Where it broke or slowed me down

1. Reference remake failed on the first try ("This run didn't finish. The agent ran into a problem."). The retry worked and gave a coherent five-slide design, sage/off-white/teal with serif type. It felt close to the reference style, though I can't say it was a perfect match.
2. On that remake's ready screen, four of the five preview thumbnails showed broken-image placeholders.
3. The ZIP export for the remake failed ("Couldn't zip. Try again"), and retrying didn't help. That's the most important bug, because export is the whole point.
4. Generation took almost 3 minutes. That's acceptable, but a progress indicator showing what step it's on would make the wait feel shorter.

Credits and trust
The first carousel cost 95 credits and the edit 22. Neither cost was shown before I pressed send. I'd strongly recommend an estimate before each run ("≈ 90 credits"). People get nervous when the meter just drops. The model picker explains the speed/credit tradeoffs, which is good, so bring that same clarity to each run.

Smaller things

• Auto picked 3:4 for the remake even though the size menu offers 4:5, 1:1 and 9:16. It's a bit unclear why.
• All media had 19 files after a few runs, some with duplicate-looking names. Grouping them per project would help.

If I were prioritising

1. Fix the failed export and the broken thumbnails.
2. Show the credit cost before each run.
3. Make the first remake run more reliable.
4. Add progress feedback during generation.

Overall a promising tool. The targeted slide edit is genuinely useful, and once exports are reliable I can see founders using it every week.

Best regards,
Erik

Tested on Web (Chrome, desktop). Signed up with Google in a few seconds and landed straight in the app with 300 credits, no card.

What I made: "Write from a topic", a 6-slide carousel on why clinics and real estate agencies in Dubai lose leads from missed calls, and how an AI receptionist fixes it. That's a real post I'd want for my own business.

1. Would I post it?
Almost. The structure is right (hook, problem, what happens next, the fix, use cases, CTA) and the copy is tight. Two things stop me. First, slide 4 added "Speaks English and Arabic" as a feature. I never said that. Inventing product claims is the one thing I can't risk on a business account. If the agent adds a claim, flag it: "I added X, keep it?" Second, it's not my brand: a clean blue healthcare look, but nothing asked me for my colours or logo first. After the slide 4 fix I'd post it with light tweaks.

2. AI slop anywhere?
Mostly no. Slide 1 is the weak spot: the illustration puts a "property for sale" sign standing on an office desk, and the desk is cut off at the bottom edge. Slides 2 to 6 look designer-made.

3. Remake: I didn't test a link remake this round; I went topic-first.

4. Single-slide edit: Yes, this worked well. I asked "Slide 4 only: remove the 'Speaks English and Arabic' line and make the headline punchier." Only slide 4 changed: the line was gone and the headline became "The fix / ZERO MISSED CALLS". The other 5 slides were untouched.

5. Stuck, waiting or confused:
On the start screen the prompt box grows as you type and the send button moves down with it. My first click landed where the button used to be and nothing happened. Enter works.

When the edit started, the big preview showed slide 6 greyed out, not slide 4, which I was editing. For a second I thought it was redoing the wrong slide. Jump the preview to the slide being edited.

Generation took about 2 minutes, the edit about 40 seconds. The progress labels ("Reading your message", "Building the slides", "Checking the slides") make the wait fine.

Cost: the carousel took 82 credits and a one-line text edit took 49 (300 to 169). A small text fix costing 60% of a whole carousel feels steep. Show the cost before I send.

6. Anything broke:
Nothing broke. Slide 6 has a big empty gap between the headline and the body text, and the CTA "Message us to try it" has no handle or link. Ask me where to send people.

Bottom line: the fastest I've gone from one sentence to a nearly postable carousel. Stop it from inventing claims, show the cost up front and let me set my brand first, and I'd use it every week.

Tested on Web. I signed up free with email (verified account, 300 credits, no card asked) and ran the builder's own test: a 6-slide carousel from the idea "why most people quit the gym".

What happened: generation took 2 minutes 44 seconds and cost 82 credits, so the balance went from 300 to 218 — one carousel is more than a quarter of the free allowance, about three carousels before paying. The result itself is good: 6 slides, 9:16, dark moody gym photos that match each other. Cover: "Most people quit the gym in 3 months. Here's why". Then Reason 1 "No plan. Just wandering", Reason 2 "Too heavy, too soon", Reason 3 "Expecting results in 2 weeks", Reason 4 "Motivation fades. No system", and a fix slide "Boring routine. Show up. Repeat." The copy reads naturally and the photos fit each reason.

Single-slide edit: I asked it to change slide 4's headline to "Expecting overnight results" on one line. It did exactly that in 44 seconds for 6 more credits, other slides unchanged, and its layout check reported no problems. That part works well.

What I would change:

1. Show the credit cost before I hit send. I only learned the carousel cost 82 credits when the balance dropped afterwards.
2. A text-only headline edit re-ran the whole pipeline, including "Make the gym photos" and a layout check of every slide, even though nothing visual changed. A text edit should edit the text in place.
3. In the editor preview, the prev/next arrows sit on top of centred headlines on slides 4-6 and cover the first words, so at a glance I could not tell whether the text was clipped or just hidden behind the arrow. Move the arrows outside the slide or hide them until hover.
4. The assistant's summary in the side panel was cut off mid-sentence ("It's in a dark, moody gy…").

Would I post it? Yes — the carousel reads well and looks cohesive enough to post under my own account. But I would want the cost shown up front, because at 82 credits a carousel, I would think twice before regenerating.