Every new engine, feature and improvement, newest first. A new engine opens to everyone automatically, on the same balance.
Upload one selfie and get an open-world game kit: a character select screen, a character sheet, a cover, your ride, a mission card and a wanted poster. The roll picks your role, city, ride and mission, so nobody else gets the same kit. Original and free to post.
Tick ready looks and your avatar gets them on its own face and voice. And in the "Cinematic scene" tab pick a scene: the avatar walks up, reacts, sits down or comes through a door, then says your line in your voice, in any of its looks.
Google's new generation, with synchronised sound born with the shot. A longer clip, stronger hold on characters and references, and clearer lips on speech.
Up to 20 seconds at 720p, from words, an opening frame or up to 10 reference images, and editing a video you already have.
Upload pictures, videos and voices to a project, sort them into folders and invite people. Your client gets a link, comments on an exact second, approves or asks for changes, and can upload files too. Add a password and an expiry if you like. Open it from "Project folder" at the top of the Genie.
Genie Edit now has a vertigo zoom, a prompt box that types itself, a behind-the-scenes window, numbers and comments landing on screen, and slots that fill up. Plus eight ready-made styles with a "Surprise me" button.
Write a change on the finished film and the rest of the edit stays as it is. You can also add clips and pictures straight from your gallery, with no download and upload.
Upload a video and the Genie splits it into scenes: the picture, what is said, and two moves that fit each one. Tap to choose or write your own note. The board is free, and you pay only when you press the edit button.
In GPT Image 2.6 choose "Animated GIF", write what moves, and you can attach a picture of a character or product and the GIF is made from it. Good for a business mascot, a reaction sticker or a product in motion. Each GIF starts at 1 credit and lands in your gallery.
OpenAI's newest generation in a tile of its own, with fast Flare and precise Sunburst, and one new choice: how much the model plans before it draws. Low for a small change, high for dense text and busy layouts. The choice adds no credits, and prices match 2.5.
Make your ID once. Tag it into anything. A character, a place, a product or a look gets a Cadabra ID, and from then on @name in any request brings it back the same in every picture and every video. In the Genie the button says "Create character ID", and "Your Cadabra IDs" under the message box opens everything you saved. Free.
A new voice engine in a tile of its own. Record one minute, clone it, and from then on every text comes out in your voice, in Hebrew, with the pace and volume you choose. A clone is 5 credits, and a read is one credit per 750 characters.
Each kind has its own colour: a character in purple, a place in teal, a product in orange and a style in pink, on the button, in the list and inside the sentence itself. The button says what you save ("Create product ID"), and under the Genie message box "Your Cadabra IDs" opens everything you saved. A tap puts it in the film.
In the Genie, on the projects list: "Copy as template" opens a new project with the same cast, worlds, sheets, brand, tone and voices, without the old film. The next campaign starts ready. Every project card now shows its cast, its place and the credits the project has used. Free.
Create a Cadabra ID for a character, a place or a product once, then type @ and pick it from the list, and it comes back the same in every shot. Cadabra writes each one a full description from all its pictures and locks its colours. Free.
Upload a clip and a picture of you in another look, and Kling 4.0 keeps the exact motion, camera and rhythm while it changes the clothes and the whole world. A walk down a hallway becomes a flight attendant on a plane, a princess in a palace or a tourist in Tokyo.
The new Kling, in a tile of its own. Several shots in one video, each with its own words and length. Up to 10 reference images you call by name in the prompt. An opening and a closing frame. Keyframes for every key moment of the film. And editing a video you already have, in words. 3 to 30 seconds at 720p or 1080p, with the price shown before you press.
Upload a 5 to 30 second video and one to four photos, and the people in the video become the people in the photos. The motion, camera, cuts and original sound stay exactly as they were. The price follows the video length and is shown before you press.
The new generation of H3: a better-looking picture, closer to what you wrote, and sound made together with the video. 480p, 768p or 1080p, 5 to 15 seconds, with a first frame, a last frame or both. Want your own soundtrack? Attach it and it is pinned to the clip. And Turbo does the same at half the price. 5 seconds at 768p is 3 credits.
Posters, logos and designs with exact lettering. Put the words that must appear in quotes and they are spelled exactly. Up to 8 images a run, at 1K or 2K for the same price. Upload a picture for a precise edit: only what you asked for changes, and everything else stays exactly as it was. One credit an image, 2 at the highest quality.
The newest Midjourney, with better aesthetics and image quality. 4 images for one credit, or a Draft of 24 sketches at the same price to pick a direction fast, then continue from any sketch with Variation or Remix. Native 2K HD and a detail level from 1 to 4 at no extra cost. Midjourney V8.1 stays exactly as it was.
In Genie Edit, tap "Add an effect from the library" and 100 edits play in front of you: speed ramps, a before and after split, a zoom on every line, an emoji on the word, a name tag, a progress bar and an end card. Search in plain words, write your own and see the result at once, and one tap puts the edit in your film. For package holders.
Finished a picture or a clip in Seedance 2.0 mini or Cadabra Morph? The next step waits right under it: the picture starts to move, another take of the clip, or a whole film with the Genie. The price is on screen before anything is made. And "Send to friends" opens a share page with your clip and a caption ready for Instagram.
Every look of your avatar waits in the avatar studio. Tap a look and the avatar is filmed in it. Tick several and continue to an avatar film in the Genie, with the avatar, the looks and the script already in. You can also put several avatars in one film, each saying its own lines in its own voice.
Write one sentence and your film gets one of ten cinematic colour looks, painted on the footage only. Or a new backdrop: crumpled paper, moving dots or VHS. Or one of twenty animation styles: neon, Y2K chrome, paper cutout, comic book, clay 3D, holographic and more. Plus three new sounds for moments of tension and drama.
One photo of you and any dance clip, and you dance it step for step with the music. The dance clip runs up to 30 seconds, with two quality levels, each priced on screen before anything.
Upload a selfie and get a 1950s horror movie poster, a vampire noble, Day of the Dead makeup, a movie zombie, a witch in a haunted forest, your face carved in a pumpkin, a Victorian ghost photograph and your pet in costume. And three films: full moon, a haunted mansion and a pumpkin field that lights up around you. No prompt to write.
Cadabra AI passed OpenAI's review and is now in ChatGPT's plugin store. Install in one click, sign in with the email that holds your credits, and tell ChatGPT what you want: a picture, a video or a whole film. The price is on screen before anything is made.
Up to 16 seconds and up to 4K, with synced speech and sound. Start from one picture, or upload up to 15 pictures of characters, clothing and places and up to 3 voices, and the prompt calls them by number. Reference pictures and voices cost nothing: you pay only for length and resolution.
Upload a clip or a selfie and pick a spell: a felt puppet, a 3D animated feature, hand-drawn anime, an action figure in its box, you at 80, you before a fairytale castle. And Morph Mix: one clip that changes style on every snap, with your voice and music underneath. No prompt to write, the price on screen before anything.
Sharper pictures and more accurate text inside them. Create from text at up to 4K, or combine up to 14 reference images into one picture: a face, a product and a place. One credit per image up to 2K.
A new caption look: the phrase builds on the empty wall above your head, word by word from your voice, the key word huge in colour. Plus: you become a comic illustration, your background is redrawn in pencil, three of you in three different poses, glass cards that each light up exactly as you say them, and a word struck out with a new one landing in its place. All from one sentence in Genie Edit.
Attach a film you liked and yours is cut to the same rhythm: we measure every cut, each shot's length, and whether it speeds up. A screen recording becomes a tutorial by itself: each step glides onto where something happened, a cursor clicks, and a numbered card says the step. You can also make an animated film with no footage at all, put three of you in one frame, and turn a moment into a paper collage moving on eights. Every effect plays live in the new catalog: livinginmovie.com/edit-effects
Tell the editor in one sentence and it draws on your film as if by hand: 68 drawings (flowers growing up both sides, an arrow, a marker under a word, cat ears and a halo on a head, stars circling a head, falling hearts, a frame of leaves, a little cat or ghost and more), in six line styles: pen, marker, chalk, neon, dashed and sketch. Lines draw themselves, then breathe, sway or spin. Ask for "stop motion" and the picture moves in jumps, a few drawings a second, while the sound plays on as usual.
ByteDance's pro image engine, in a tile of its own. Create from text, or add up to 10 pictures and change only what you asked for, leaving the rest as it is. Send it a layer from Split into layers and it comes back changed, still transparent and ready to go back into the composition. One credit per image.
Upload one picture and it comes apart into a base plus up to 16 transparent layers: every person, object and line of text on its own, with whatever hid behind them filled in. Move, resize, hide and reorder any layer, send one to be edited, and download everything as a ZIP for Photoshop, Figma or After Effects. You pay only for the layers that came out, from one credit.
Create a picture from text at up to 4K, or edit one you have and change only what you asked for, leaving the rest as it is. Combine up to 10 reference images into one picture, a face, a product and a room, with no charge for the references. One credit per image up to 2K.
Upload a long recording and every lesson comes out clean: pauses, "um", a word said twice and a sentence started over are gone, and the last attempt stays. On a screen recording a pause is cut only where the screen stands still, so typing, clicking and waiting for a result stay. And every lesson plays at the same volume, the level the platforms and phones play at. No extra credits.
Upload a video to Genie Edit and write what you want, on which word: "on the word now, shatter the screen". The screen cracks and shatters, the whole film flips, time freezes, a grey open slams into colour, a 3D title flies in, stamps land, a price drops to zero, a comment types itself and a message arrives, a follower counter climbs, a follow button gets clicked, a hologram reveal, an infinite zoom into the screen, a spinning cube, and a rewind to close the loop. And now you yourself: step out of the screen, move to space, underwater or a neon city while you keep talking, grow giant, or break into pixels and come back. You can also ask for the viral-reel layout: you in a framed card, a fixed headline on top and a call to action below. Every effect has its own sound and lands exactly on its word. And the captions: karaoke that lights each spoken word, under the chin and never on the face, English inside Hebrew in the right order, and 10 styles to choose from.
How an animation studio works, as one road. Design a character from words or a photo, fix it by hand, and lock one look: 3D, clay, anime, storybook or live action. Give each scene its movement from whatever is easiest: a grey Blender blockout, your own sketch, a phone clip or words alone. Every scene comes out with the same character and look, and every step is priced and approved first. The video is only one example: with Blender you plan any movement, from a car chase or a fight to a dance or a drone shot over a city.
Turn on "Smart draft" and test the shot at the 480p price. Like the take? One click renders that very same shot at 1080p: the same composition, motion and sound. Not an upscale and not a new take. Valid for 24 hours.
Pick a voice for the character once and it stays with it for every film. Clone your own voice right inside the studio, or pick any voice from the library. And a new door, Eleven v4, lets the character perform the text: a whisper, a laugh, excitement. Same price.
A narration in one voice or a scene of up to four speakers, directed inside the text: a whisper, a laugh, a sigh, a sound effect or an accent, one tap on a tag. Every take sounds like the same speaker, and you can redo a single line. 90+ languages, Hebrew included. One credit per 750 characters.
Plan the camera move in Blender for free, in grey, and Seedance 2.5 films the real scene along that exact path. No trial and error and no paying for every try: perfect it for free, then pay once for a shot that does what you asked.
A narration in one voice or a two-speaker scene line by line, with a direction for every line and a laugh or a sigh inside the text. Pick from the voice library or design a voice from a description, and fix one line without redoing the scene. About one credit per minute of speech.
The newest Suno joins the studio, beside the one that was already there. Three things it does that the older one cannot: you say how long the song should run, anywhere from 15 seconds to six minutes, and it composes a real piece at that length instead of handing back a song you have to cut. You get two different takes of the same brief, and both are yours. And it sings Hebrew, so you can write the words in Hebrew and hear them sung exactly as written. All of it for one credit, at any length. Once the song exists you can carry it on for longer, split it into separate instrument tracks, or turn it into a music video. Two things come with the song at no extra credit: a studio-quality WAV, and the lyrics with a timestamp on every word, which is what accurate karaoke captions are made from.
A new engine in the studio, and the cheapest route we have to a long take. Three modes: from text, from an image (a first frame, and an end frame if you want one), and a reference mode that takes up to 10 pictures, 5 clips and 5 sounds at once, plus a document or a link to a web page. The prompt calls the material by number, for example "the character in Image 1 walks into the room from Video 1". The length is any whole number between 2 and 30 seconds with no fixed grid, so a 7-second beat comes out 7 seconds. Sound is born in the same pass and costs the same whether it is on or off. One thing to know: a reference clip adds its own length to the bill, while pictures, sounds and documents are free.
GPT Image 2.5 now shows the picture being drawn instead of a grey box, and after the first fix the engine already holds your image: write only what changes, "now realistic", "lose the hat", with no re-upload and no re-describing the whole scene. Prices on this tool came down at the same time: a 2K high-quality image went from 10 credits to 5, and on the precise model from 20 to 8.
GPT Image 2.5 now has a brush: upload a picture, paint the area you want changed, and write what goes there. What you did not paint is kept: same face, same product, same light. The engine treats the mark as strong guidance, so paint a little wider than the area itself. Two new quality rungs for a final frame, a background choice (auto, opaque or transparent) and a file format choice of PNG, JPEG or WEBP with a compression level came with it. The price is shown before every run.
Two new image models, and you choose between them: Flare is the fast one for everyday work and volume. Sunburst is the precise one, for the moment you upload a real photo of yourself or your product, ask to change only the jacket or only the background, and get exactly that back, with the same face and the same label. Hebrew inside the picture comes out right, and a real transparent background gives you an asset that drops straight into a site. The price is shown before every run, as always.
A gift for the holiday: drop one photo and get a filmed greeting starring you, drawn from eleven scenarios at random so no two people get the same one. The frame is 1 credit and you see it before paying for the film; the five second animation is 5 more.
Five routes in one engine: text, image, first-and-last frame, references, and editing or extending a clip you already have. Synchronised sound on every route, up to 4K. Five seconds at 720p for six credits.
The cheap engine for finding a shot. Five seconds for three credits instead of sixteen, so ten ideas cost what one used to.
Upload a clip and change what is inside it, or continue it with unbroken motion for up to thirty more seconds.
Home · Pricing · What's new · All engines · All effects · Help center · Lessons · Meital Sela · AI video · AI images · AI voice · AI music · Image to video · What it costs · Credits, not a subscription · Blog · Connect to Claude and ChatGPT · Presets · Privacy · Terms
Cadabra AI · Meital Sela · Rehovot, Israel