The group-chat reply
A photo everyone already knows, plus four words, sent before the conversation has moved on.
Your photo in · A bubble that fits the frame · No watermark
Put a comic bubble on your own photo, with text that actually reads — online, free to start, and nothing stamped across the export.
A real run through the model this page uses. Drag the divider: the bubble arrives in the empty upper left, the tail points back at the cat, and the caption is readable at a glance — which is the whole test. Every output is AI-generated, so your result varies with your photo and how long your line is — treat this as a guide to what the tool can do, not a fixed result you will get back exactly.


Drag to compare before and after
It takes a photo you already have and adds a comic-style bubble carrying whatever line you want, so the subject appears to be saying it. The format is old — it is the oldest joke on the internet — but the tools for it split into two disappointing halves, and that split is the reason this page exists.
On one side are the classic meme sites: fast and free, and they hand you back an image with their brand burned into the corner. On the other are real editors, where you place a shape, pick a font, size the tail and nudge everything into position — five minutes of work for a joke with a thirty-second shelf life. Neither is wrong; both are more than the job deserves.
The bubble pass runs on the Nano Banana image model, in the browser, with free credits to start. Upload the photo — the flow defaults to image-to-image — type the line you want in the bubble, and the credit estimate updates before you commit; a failed run is refunded automatically unless it broke the content policy.
Two things decide whether it comes back usable. The first is empty space: a bubble needs somewhere to sit, so a photo with a clear wall or sky in one corner works and a tight crop full of face does not. The second is length — a short line renders cleanly, and a paragraph is where the lettering starts to break down, because rendering text is still the hardest thing you can ask an image model to do.
Creative engine
Upload a photo with some empty space, type the line, and the credit estimate updates before you generate.
Bold black comic text inside a clean white bubble with a crisp outline — readable at the size people actually view it.
Runs in the browser with free credits — no editor, no layer panel, no font hunting before you see a result.
It is one pass over a photo you already have, the same as the beams in AI pet laser eyes, the exaggeration in the caricature generator, or the felt rebuild in the muppet generator.
From your own photo to a posted joke, with the cost shown before you commit.
Leave a corner of plain wall, sky or floor free. That gap is where the bubble goes, and a frame packed edge to edge leaves nowhere to put it.
Keep it short — a handful of words renders crisply. Say where you want the bubble and who the tail should point at if the photo has more than one subject.
Generate and post it. A failed run is refunded unless it broke the content policy, so rewording the punchline costs nothing but the wait.
People replying in a group chat, people running an account, people making a card.
A photo everyone already knows, plus four words, sent before the conversation has moved on.
Give the recurring character a voice without paying an editor's five minutes for every post.
An inside joke printed on a birthday card, or the one slide in the deck that wakes the room up.
Common questions about captioning a photo with Kavel.
No. That is the specific complaint that sends people looking for another option — several of the best-known free meme sites fund themselves by branding the corner of every export. Nothing is stamped onto what you download here.
For a short line, yes — the demo on this page is a real run and the caption reads cleanly. Text is genuinely the hardest thing to ask an image model for, so the honest version of this answer is: keep it short and it holds up, write a paragraph and letters start to go wrong. If a run comes back with a mangled word, generate again.
In the emptiest part of the frame, with the tail pointing back toward the subject. You can direct it — say upper left, or say which person should be speaking — and it will follow that. What it cannot do is invent space in a photo that has none.
Yes, if the photo has room for them and there are separate subjects to attach them to. Two people having a two-line exchange works. Six bubbles on a group photo is asking for trouble; build that one up over a couple of runs instead.
Yes — describe it. A thought bubble with trailing circles, a jagged shout burst, a rounded manga bubble, a black bubble with white text. The default is the plain white comic bubble with a black outline because it reads at any size.
Any photo with somewhere to put the bubble. Portraits, pets, screenshots and product shots all work. Very dark images can make a white bubble sit oddly, and a heavy crop that fills every pixel with subject leaves the bubble nowhere to go but over a face.
The estimate shows before you generate and moves with the format and resolution you pick, so you know the cost first. Failed runs are refunded automatically unless they broke the content policy, and new accounts start with free credits — enough to caption a few photos before spending anything.
Technically yes, and it is worth a second of thought first. Putting words in a real person's mouth is the one way this format goes wrong — a friend who will laugh is fine, a stranger or a public figure is a different thing. The content policy applies to what you generate here either way.
One credit pool covers Nano Banana images and Seedance and Veo video. The cost shows before every run, and failed jobs are not charged. Use a subscription for ongoing work, or a one-time pack when you just need to top up.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
One-time top-ups — buy extra credits any time you run low.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “KAVEL AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for.
Powered by
Use one balance across every supported image and video model — ten of them today, listed below — and check the credit cost before each request.
Video models
Image models
Model credit guide
You're only charged for successful generations. The exact estimate in the generator varies by model, length, resolution, audio, and number of images.
| Type | Model | Credit cost |
|---|---|---|
| Video | Seedance 2.0 | 5s 720p image-to-video ≈ 188 credits; text-to-video ≈ 308 credits. Scales with resolution and length. |
| Video | Seedance 2 Fast | Faster and lower cost. 5s 720p text-to-video ≈ 248 credits. |
| Video | Seedance 2 Mini | The cheapest Seedance tier. 5s 720p ≈ 154 credits, 5s 480p ≈ 72 credits. |
| Video | Seedance 1.5 Pro | Audio doubles the rate. 5s 720p ≈ 27 credits silent, ≈ 53 with audio; 1080p ≈ 57 and ≈ 113. |
| Video | Veo 3.1 | Billed per video, not per second (Lite tier). About 45 credits at 720p and 53 at 1080p. |
| Video | Kling 3.0 | Audio raises the rate. 5s 720p ≈ 105 credits silent, ≈ 150 with audio; 1080p ≈ 135 and ≈ 203. |
| Video | MiniMax H3 | Fixed 2K, no resolution ladder. Priced per second — a 5s clip ≈ 158 credits. |
| Image | Nano Banana 2 | Generate or edit from text and images. About 8 credits per 1K image, 12 at 2K, 18 at 4K. |
| Image | Nano Banana Pro | Consistent run times across generations. About 12 credits per 1K or 2K image, 21 at 4K. |
| Image | Nano Banana 2 Lite | Faster, simpler variant. A flat 8 credits per image at every resolution. Start here to test. |
| Image | GPT Image 2 | The lowest-cost image model here. About 3 credits per 1K image, 6 at 2K, 12 at 4K. |
| Image | Seedream 5.0 Lite | Flat pricing — about 8 credits per image whether you generate or edit. |
The punchline goes into the bubble as typed, and nothing is stamped across the export.