OpenAI has extended ChatGPT into a new kind of making: users can now ask it for animated GIFs, a small moving picture drawn from words, as part of a new image model called Images 2.6. The release also lets the model begin from a picture a person supplies, and it offers faster versions to developers who build image tools for others. Much of what matters about it, how well it performs and whether it outpaces its predecessor or its rivals, is still waiting to be measured by people who have not yet had their hands on it.
ChatGPT Adds Animated GIF Creation Through Its Images 2.6 Model
ChatGPT can now make animated GIFs.
Mark
So what actually changed in ChatGPT this week?
Mimi
It can now produce animated GIFs. That arrives with Images 2.6, along with support for reference images and faster models in the API.
Mark
Why would a GIF maker matter? People already have GIFs.
Mimi
It moves image generation from still pictures to short animated clips, inside a tool people already use. That's the shift the headlines are pointing at.
Luke
Careful, though. What we have are headlines, not the articles. We know the features exist. We don't know how long the GIFs can be or what limits apply.
Mark
Outlets are already testing it, right?
Mimi
Yes. Kingy AI compared Images 2.6 with 2.5 on cost and speed, and RuntimeWire set it against Nano Banana 2.1.
Luke
But we haven't seen the results of either. Anyone saying 2.6 is faster or cheaper is getting ahead of what's in front of us.
Mark
And the API side?
Mimi
Faster models for developers, which is how other products plug in image generation. Specific models and prices aren't in what we have.
Luke
Also worth noting Decrypt called the model unannounced, so the rollout may be uneven. I'd wait for independent benchmarks before judging it.
Le Pouls
- ChatGPT can now turn a request into an animated GIF, a capability that arrived quickly and drew attention before the full model was formally presented.
- Reference-image support lets users hand the model a starting picture, though how many images are allowed and what limits apply remain unclear.
- Outlets are racing to test Images 2.6 against Images 2.5 on cost and speed, but no verdict has yet settled whether the upgrade is actually cheaper or quicker.
- A head-to-head with the rival system Nano Banana is already being framed, with no clean numerical score yet available to decide the contest.
OpenAI has extended ChatGPT into a new kind of making: users can now ask it for animated GIFs, a small moving picture drawn from words, as part of a new image model called Images 2.6. The release also lets the model begin from a picture a person supplies, and it offers faster versions to developers who build image tools for others. Much of what matters about it, how well it performs and whether it outpaces its predecessor or its rivals, is still waiting to be measured by people who have not yet had their hands on it.
ChatGPT can now make animated GIFs. That is the headline of an update OpenAI is rolling out alongside a new image model, Images 2.6, and it is the part of the release that most coverage has seized on first.
The picture is still forming. Reporting from several outlets points to three additions: GIF creation, support for reference images that let a user supply a starting picture, and faster models offered through the API that outside companies use to build image generation into their own products. How many images can be supplied, which API variants are faster, and what they cost are not yet spelled out.
Testing began almost at once. Kingy AI has set Images 2.6 against 2.5 on cost and speed, and RuntimeWire has pitted it against the rival Nano Banana, though its headline suggests there was no clean score to compare. The results are not yet available here, so no verdict on performance can be given.
What can be said firmly is limited: the model is out, GIFs are generated, reference images are supported, and API models are faster. Everything about how well it performs is still to come, and independent benchmarks will tell the rest.
ChatGPT can now make animated GIFs. That is the headline of an update OpenAI is rolling out alongside a new image model, Images 2.6, and it is the part of the release that most coverage has picked up first.
The source material here is thin: a cluster of headlines from different outlets rather than a full article. What those headlines establish is the shape of the release. The Verge reports that OpenAI has launched ChatGPT Images 2.6 with three additions: GIF creation, support for reference images, and faster models available through the API.
The GIF feature is the one the reporting leads with. A user can ask ChatGPT for an animated clip, and the model behind it is Images 2.6. Decrypt describes the GIF maker as powered by what it calls an unannounced Images 2.6 model, and its framing suggests the capability surfaced in ChatGPT before OpenAI had fully presented the model itself. That detail comes from a headline alone, so how the rollout was staged is not spelled out.
Reference-image support is the second addition. It means the model can take an image supplied by the user as a starting point, rather than working only from a text description. The headlines do not say how many images can be supplied or what limits apply.
The third change is aimed at developers. OpenAI is offering faster models through its API, the route by which outside companies build image generation into their own products. The headlines do not name the specific model variants or give pricing.
Testing and comparison have begun almost at once. Kingy AI published a piece titled "ChatGPT GIFs Tested: Images 2.6 vs. 2.5, Cost and Speed," which sets the new model against its predecessor on how much it costs and how quickly it works. The results are not included in the material available here, so no verdict on whether 2.6 is cheaper or faster can be given.
Rivals are part of the conversation too. RuntimeWire has a piece headlined "GPT-Image-2.6 vs Nano Banana 2.1: No Score to Compare," pitting the model against a competing image system. The title hints that the author found no clean numerical scoring to rely on, but the substance of that comparison is not available either.
So what can be said firmly is limited: OpenAI has released Images 2.6, ChatGPT now generates animated GIFs, reference images are supported, and API models are faster. Everything about how well it performs is still to come. Watch for independent benchmarks and head-to-head tests against Nano Banana as the feature reaches more users.
Envie de l'histoire complète ?
Lire l'original sur Google News ↗
Technology