ChatGPT Can Finally Generate GIFs: Images 2.6 Launches with a Surprise Upgrade

  It's the first time in the history of the world.

Get ready, OpenAI is making a big deal!

Just recently, a hotshot, ChatGPT Images 2.6 launched a massive upgrade, with the graphics speeding up by 50 percent.

This is the first time, and the first OpenAI initiative, that a single click of an AI-generated image directly generates a GIF.

Not only that, GPT Images 2.6 also supports step-by-step editing, playback adjustment, and synchronization of live images.

The sand-sculpted horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse-drawn horse.

Today, the two new APIs, Flare and Sunburst, are officially available.

The former charted the fastest in 6 seconds, while the latter dominated high-quality and editing consistency, and supported SVG output, local refinement.

Fortunately, the next generation of GPT-6.2 Astra and Sol are also on the Google platform, and many people in the circle are betting: it will be officially announced tomorrow!

ChatGPT is finally able to make GIFs, and a picture moves directly.

UIS1, OpenAI on the other hand, is an upgrade to GPT Images 2.6, which is mainly a utility.

In the past, you had an AI to draw a kitten, and what you get is usually a static image.

If you want to make a kitten dance, blink, wave its tail, you need additional animation software or other generating tools.

This time, ChatGPT took over the entire process directly.

He wanted a photo to move and threw a phrase at ChatGPT: "Can you make it into a GIF?"

In the blink of an eye, the protagonist in the photo is instantly "alive".

The cat with the little knitted hat turned around, and the frogs under the disc lamp ball, the big duck on the top of the pink car, and the monkey on the top of the pink car, all jumped with the beat.

Most importantly, it can continue to be modified after generation, adding a laser to the blue bird and adjusting to twice the speed.

The main character moves a mouthpiece and changes the plot to magic. The most interesting part of GPT Images 2.6 is that it can also be "edited step by step".

Photos of pets taken with their hands on their phones can also be turned into dance moves.

Putting pictures of turtles, lizards and lizards in your home into a dialog box, you turn your eyes to a baby bear dancing in a bow tie, a baby bear blinking in a bouquet of flowers, and a baby bear singing with a microphone.

Even a hand-painted picture of an octopus, transmitted to ChatGPT, can instantly grow brown feathers, waving wide-eyed greetings.

It's also a good idea to use it as a daily chat bag.

A Q-version worker on the left staring at the collapse red line crying, and on the right staring at the rising green line cheering for money.

We've been playing a game, and we've been playing a game.

After all, it was the first time that ChatGPT was able to move, and we naturally couldn't stand it.

First, we're going to use the super-IP "super beans" that we've just unveiled, and then, in a word, we're going to create a bit of a bit of a bit of a "crazy" code map at the secondary level.

In addition, it is possible to make handmade expressions, characteristics and hairstyle details a bit smoothly.

Another IP superhero manga, which is a true science fiction hero, has a visual effect.

We had ChatGPT play a "backstabbing mouse" out of the loop.

The two little claws go to the back of the knot, staring silently out the window, and the melancholy is directly pinched dead.

The Blue Big Fish, also alive on ChatGPT, is cooking and waving to the programmers.

Continuing to lock in the core people of the big fish, ChatGPT generates a second of "deep thinking" and reverses it to "a few more bowls".

Next, make it into a motion, a whole set of micro-faces and small movements, with no knitting ends, eyelashes, tails.

It's also the place where you can play with classic names like Star Wars and Dragon Ball.

As soon as the new feature was released, X netizens were playing for the first time.

Tibo has the reset key.

Professor Derya Unutmaz of the Jackson Laboratory hand-generated an animation of the character Cartman, who laughed wildly in his head, with the character's theology quite restored.

The tech blogger TestingCatalog would be even more playful, with a line of cloaked kids watching the endless loop of "Daily AI News" in the newspapers, looking at the magic.

In addition to the moves, the "Save Reference Chart" function has also been arranged.

Save photos of yourself, your family, friends, or pets, and then, no matter what new chat box you open, you can use direct@ at any time, both for savings and fun.

4.84 seconds to the top of the chart, 50% faster than life.

It's easy to say that the front end is so smooth, and the back end is lined with two new models that are simultaneously landing in the API.

GPT-Image-2.6 Flare 

GPT-Image-2.6 Sunburst

OpenAI, by the way, is a muscle, and images generated weekly on ChatGPT, Codex and APIs by users around the world have broken through.4 billionDay by day500 million copies。

On such a large scale, version 2.6 increased the overall graphics speed by 50%.

I'm not sure.FlareDesigned for high-speed graphics and print design, low-quality grades can be rolled in less than 6 seconds.

A luxury travel advertising poster was also screened, with the previous generation 2.5 Flare taking 7.08 seconds, while 2.6 Flare was used only.4.84 secondsAfter completing the rendering, the speed increased by 46%.

Changing interaction has also led to a major upgrade supporting red line graffiti.

Draw a fork, draw an arrow, and Image-2.6 will accurately capture the image and automatically erase the trace.

The photo below shows the transformation of the living room, which is clear at a glance.

Another dominant modelSunburstThe main focus is on high-detail consistency with multi-sequence editing.

The three steps of the red-haired female knight repeatedly altered the test, and in the last picture, the female knight's five senses, theology and sword-bearing posture are always consistent.

It's worth noting that the two new models took the number one spot on DesignArena's third-party blind list as soon as they debuted.

SunburstOne climbed to the top of the overall picture arena rankings and took home the product and landscape singles championship;FlareIn the first three revenue bags, image, architecture, and font, the duo refreshed SOTA's quality and speed.

Nine-Mira straight out GIFs, originally exporting SVG vector graphs, without adding value

This time, OpenAI has written GIF animation, SVG vector graphics and structural complexity controls into the native API.

It's very simple for developers to use.

If you set the output format to gif in the API, you can freely specify the number of gifs between 2 and 64 bits per millisecond.

At the bottom of the model, a sequence of sequential "nine-dimensional" movements is executed, followed by a sequence of automatically aligned slices, combined into a seamless cycle of championship movements or snowflakes.

The early testing team Town had already used this capability to replace their own AI office assistant Monty with live-action typewriters waiting for the animation.

For the first time ever, the new model also supports native export for front end engineers and UI designers.SVG vector format。

The geometric animal icons and interface buttons that are generated have full XML vector paths, and drag-and-drop design software can be used to infinitely enlarge and disassemble the colors.

OpenAI researcher Gabriel Chua says that images 2.6 can also be better at controlling image complexity.

What's more, the major version of the upgrade is a complete "add-on no-add-on".

FlarewithSunburstThe API pricing is exactly the same as the previous generation 2.5, with text input costing only $5 per million tokens, image input costing $8, image output costing $30, and graphite marking and complexity control not charging any additional fees.

Currently, ChatGPT Images 2.6 is open to all subscription-level ChatGPT and Codex web, desktop and mobile users worldwide.

What's more, a wave of GIFs is already on its way.