Today, ChatGPT can finally generate GIFs! Images 2.6 has launched unexpectedly
Source: [New Intelligence](https://mp.weixin.qq.com/s/_K3CGhTQV9qqDDRJoPuDcw)
Get ready, OpenAI is about to make a big move!
Just now, as a warm-up, ChatGPT Images 2.6 has received a major upgrade, with image generation speed skyrocketing by 50%.
This is also the first time, and a first for OpenAI, that AI-generated images can be directly output as GIFs with one click.
Not only that, GPT Images 2.6 also supports frame-by-frame editing and playback speed adjustments, with the reference image feature launching simultaneously.
With a playful rooster striking a pose, after speeding up, it directly performs a chaotic dance.
Today, these two new API image generation models, Flare and Sunburst, are officially online.
The former can generate images in as fast as 6 seconds, while the latter focuses on high quality and editing consistency, and supports SVG export and local refinement.
Coincidentally, the next generation GPT-6.2 Astra and Sol have also appeared on the Google platform, and many in the industry are betting that an official announcement will come as soon as tomorrow!
### ChatGPT can finally make GIFs, bringing images to life
UIS1, the upgrade of GPT Images 2.6 from OpenAI focuses on practicality.
In the past, if you asked the AI to draw a kitten, you would usually receive a static image.
If you wanted the kitten to dance, blink, or wag its tail, you would need to use additional animation software or other generation tools.
This time, ChatGPT has taken over the entire process.
Images are no longer monotonous; if you want a photo to come to life, just tell ChatGPT, "Can you make it a GIF?"
In an instant, the subject of the photo comes to life.
A cat wearing a daisy knit hat starts twisting its body, while a frog under a disco ball, a proud rooster strutting, and a rabbit on a pink convertible all dance to the beat.
The key point is that after generation, you can continue to modify it, adding lasers to the bluebird and adjusting the speed to double.
With just a few words, you can transform the image. The most interesting part of GPT Images 2.6 is that it also allows for "frame-by-frame editing."
Even a pet photo taken casually on a phone can be turned into a dancing GIF with one click.
Upload photos of your turtle, hamster, and parrot into the chat box, and they instantly become a little bear dancing in a pleated skirt, a Pomeranian blinking among rose petals, and a raccoon singing with a microphone.
Even a hand-drawn sketch of an owl can instantly sprout brown feathers and wave hello when sent to ChatGPT.
Using it to create daily chat stickers is also very convenient.
A chibi worker, with one on the left crying at a plummeting red line and the other on the right cheering and throwing money at a rising green line, these scenes are automatically stitched together into a looping GIF after ChatGPT generates 6 frames of action.
### Transforming meme images, we had a go at it
Since it's the first time ChatGPT can create lively images, we couldn't resist trying it out ourselves.
First, we tested it with our newly launched super IP "Super Dou Dou," and with one sentence, it instantly generated a cute image of a little Pomeranian typing furiously.
Moreover, we can also create stickers, with character traits and hair color details perfectly intact.
Another IP, Super Pony, immediately gives off a strong sci-fi male lead vibe upon appearance.
We then had ChatGPT generate a "backhanded possum" meme image.
With its little paws clasped behind, gazing quietly out the window, this melancholic vibe was perfectly captured.
The blue fat fish also comes to life under ChatGPT's pen, eating while waving at the programmer.
Continuing to lock in the core character of the fat fish, we had ChatGPT generate an image that flips from "deep in thought" to "give me a few more bowls" in the previous second.
Next, we had it create a GIF, seamlessly connecting a series of micro-expressions and small actions like holding a bowl, blinking, and wagging its tail.
Classic scenes from films like "Kung Fu Hustle" and "Empresses in the Palace" can also be played around with.
As soon as the new feature was released, X users were quick to have fun with it.
Tibo has already hit the reset button.
Professor Derya Unutmaz from Jackson Laboratory casually generated a GIF of the classic character Cartman from "South Park" laughing while slapping his thigh, with the character's expression remarkably restored.
Tech blogger TestingCatalog got creative, making an infinite loop GIF of a row of little people wearing straw hats reading the "Daily AI News," which looks incredibly amusing.
In addition to GIFs, the highly anticipated "Save Reference Image" feature has also been arranged.
You can save photos of yourself, family, friends, or pets once, and in any new chat window, you can directly @ to pull them out for use anytime, making it both convenient and fun.
### 4.84 seconds for rapid image generation, image generation speed increased by 50%
It must be said that the smooth experience on the front end relies on these two new models that have launched simultaneously with the API------
GPT-Image-2.6 Flare
GPT-Image-2.6 Sunburst
OpenAI has flexed its muscles, with global users generating over 4 billion images weekly on ChatGPT, Codex, and the API, averaging more than 500 million images per day.
With such a massive scale, version 2.6 has improved the overall image generation speed by 50%.
The Flare model is specifically designed for rapid image generation and layout design, with the lowest quality setting able to deliver results in less than 6 seconds.
In a simultaneous rendering of a luxurious travel advertisement poster, the previous generation 2.5 Flare took 7.08 seconds, while 2.6 Flare completed the rendering in just 4.84 seconds, a speed increase of 46%.
The image editing interaction has also received a significant upgrade, now supporting redline doodles.
Circle, cross, or draw an arrow, Image-2.6 can accurately understand the intent, and will automatically erase any traces after generating the image.
The following living room renovation image is clear at a glance. The completed image on the right is seamless, with even the sunlight casting shadows on the wooden floor transitioning quite naturally.
Another main model, Sunburst, focuses on high detail and consistency in multi-round continuous editing.
In a three-step continuous image editing test of a red-haired female knight, the facial features, expressions, and sword-holding posture remained consistent in the final image.
It is worth mentioning that both new models achieved multiple firsts on the third-party blind test leaderboard DesignArena as soon as they were launched.
Sunburst topped the overall image arena leaderboard and won the individual championships for product and landscape; Flare secured first place in portrait, architecture, and typography, both models setting new standards for quality and speed.
### Instant GIFs in a grid, native SVG vector export, more quantity without extra cost
This time, OpenAI has integrated GIF animations, SVG vector graphics, and control over compositional complexity directly into the native API.
For developers, using it is very simple.
Just set the output format to gif in the API, and you can freely specify the number of frames and the duration per frame between 2 to 64 frames.
The model first generates a coherent "grid" sequence frame sprite, then automatically segments, aligns, and composes it into a seamless loop of koi swimming or a lantern moving in a snowy night.
The early testing team Town has already used this capability to give their AI office assistant Monty a lively typing wait animation.
Addressing the pain points of front-end engineers and UI designers, the new model also supports native SVG vector format export for the first time.
The generated geometric animal icons and interface buttons come with complete XML vector paths, allowing for infinite scaling without loss of quality and easy color modification in design software.
OpenAI researcher Gabriel Chua stated that Images 2.6 can also better control image complexity.
What’s even better is that this major version upgrade has achieved a complete "more quantity without extra cost."
The API pricing for Flare and Sunburst remains the same as the previous generation 2.5, with text input costing only $5 per million tokens, image input $8, image output $30, and no additional fees for doodle annotations and complexity control.
Currently, ChatGPT Images 2.6 is fully open to all subscription tiers of ChatGPT and Codex users on web, desktop, and mobile platforms worldwide.
What are you waiting for? A wave of GIFs is already on the way.
References:
https://openai.com/index/introducing-images-2-6/
https://x.com/OpenAI/status/2118124944928432327
https://developers.openai.com/api/docs/guides/image-generation