Smartphone resting on a wooden surface with a glowing strip of photos winding from its screen, set against a sunset background.

Quick answer

ChatGPT Images 2.6 and Gemini Omni 1.2 Flash are the two updates most small teams can try this week. Matt Wolfe’s round-up covers about 14 AI updates, from animated GIFs in ChatGPT and free 15-second videos in Google Vids to open-weight image and video models, Meta Muse voice calls and Grok Bot phone calls. Most of the other items are either niche (cyber-security models, huge open weights) or early. The practical move is to pick one shipped feature, test it on a low-risk task, and skip the rest until it is stable.

Key takeaways

  • ChatGPT Images 2.6 can create and edit animated GIFs, and Matt’s still-image test finished in about 20 seconds.
  • Google Vids lets any Google account generate up to three free 15-second Gemini Omni 1.2 Flash videos per month.
  • Open-weight options arrived too: Pruna PH3 Image (~55 GB weights) and Pika H3 Flex (~220 GB), both with free Hugging Face demos.
  • Agents are talking: Meta Muse now supports voice and video calls; Grok Bot can place phone calls on your behalf.
  • DoorDash and Uber say almost nobody uses AI agents for deliveries or rides yet, so agent “killer apps” are still early.
  • Your action: try one free or already-paid feature this week, and leave the multi-terabyte models alone unless you have serious hardware.

Some AI weeks feel like a product launch. This one, as Matt Wolfe puts it, is closer to a rapid-fire list: lots of smaller updates, few earth-shakers. That is useful if you are trying to stay current without rewriting your whole stack.

Below we summarise the updates that matter for people who use AI for content, marketing and everyday work, and we flag what is still early or hardware-heavy. Credit for the round-up goes to Matt Wolfe (@mreflow); the judgements here are ours.

What is new in ChatGPT Images 2.6?

OpenAI’s Images 2.6 is available in ChatGPT and the API. Matt’s takeaway is that it is faster and better at “professional” visuals, with one headline feature: animated GIF generation and editing.

Claim in the videoWhat Matt showedPractical note
Faster generationStill image of a San Diego skyline in about 20.4 secondsNoticeable speed-up versus earlier ChatGPT image models, per the demo
Animated GIFsMonkey roller-skating GIF, then edited with a baseball hat and a street backgroundEdits can stack; each take was roughly 20–30 seconds
Where it livesChatGPT and the APIUseful if you already pay for ChatGPT

If you want a step-by-step for marketing GIFs (sizes, email support, accessibility), we already covered that in how to make an animated GIF with AI.

💡 Tip: Treat GIF quality as a draft. Check looping, text readability and file size before you drop one into an email campaign.

🛠️ Practical guides

Ready to put this into practice? These step-by-step guides answer two common follow-up questions:

What is Pruna’s PH3 Image model?

Pruna AI released PH3 Image, an open-weight image model derived from MiniMax’s H3 video model. Matt highlights that you can download, fine-tune and run it yourself if you have the hardware.

  • Weights are about 55 GB for the main download, with smaller quantised variants.
  • At 8 megapixels, Matt cites about 56 GB VRAM for 16-bit and about 21 GB for a 4-bit build — so a 4090 or 5090 might handle the smaller one.
  • There is a free Hugging Face Space for testing. In Matt’s San Diego skyline prompt, quality lagged ChatGPT Images 2.6 badly.

⚠️ Warning: Open-weight does not mean “easy on a laptop”. Confirm VRAM, licence and privacy before you download anything that large.

What changed in Gemini Omni 1.2 Flash?

Gemini Omni 1.2 Flash is Google’s updated video model. The practical changes Matt calls out:

  • Clips up to 15 seconds (previously about 10).
  • Price cut of about 30% from Omni 1.1, to about $0.07 per second at 720p (so ~$1.05 for 15 seconds).
  • Better reference adherence, editing and speech, per Google’s claims as shown in the video.
  • Available in the Gemini API, Flow and Google Vids.
  • Free tier: any Google account can generate up to three 15-second videos per month in Google Vids at vids.new.

Matt’s free test used his own face as an avatar, a blue shirt reference and a 15-second prompt. Generation took about a minute. Our guide on making a free AI video in Google Vids walks through a simple first try.

What is Pika H3 Flex?

Pika H3 Flex is an open-weight video model. Matt says it can go up to 4K, generate with soundtracks, and (on serious hardware) produce 15-second clips in as little as about 7 seconds. The weights are about 220 GB, and the “5-second” performance Matt cites used an H100 with 80 GB of VRAM.

There is a Hugging Face Space for free tests. Matt’s “Will Smith eating spaghetti” prompt was blocked for GPU duration at 15 seconds, then worked at 8 seconds. Treat demo queues as busy and rate-limited.

Can you now voice-call Meta Muse?

Meta Muse now supports voice chats and video calls. In Matt’s demo, Muse switched voices on request, pulled Instagram content ideas from Content Studio, and listed connected Gmail and calendar help. The avatar is optional; the useful change is speaking instead of typing.

We covered Muse’s early lessons in Meta Muse AI assistant lessons. Voice does not change the privacy basics: only connect accounts you are happy an assistant can read.

Can Grok Bot make phone calls for you?

Grok Bot can now place phone calls for you. Matt tested a call to his own number, posing as a comic-book shop asking about hours. The assistant introduced itself as a virtual assistant on a recorded line and sounded usable.

Matt’s honest caveat: many people hang up on anything that sounds like a robocall. Our guide on testing AI agent phone calls covers a safe trial that does not burn a real customer relationship.

What else made the rapid-fire list?

UpdateWhat the video saysWho it is for
GPT-6 CyberCyber-security model; access limited to security professionalsSecurity teams only
Shared ChatGPT / Codex usageUsage allowances blended across ChatGPT and CodexCodex + ChatGPT power users
Kimi K3.1 Pro weightsOpen weights now on Hugging Face; ~1.56 TB, 2.8T parametersLabs with serious hardware
Claude bank connectRead-only bank link via Plaid on Pro/Max/Ultra; US roll-outUS Claude subscribers willing to share finance data
Selendis (Steve Aoki)AI video editor in early access; learns taste from editsCreators waiting for invites
Netflix AI Discover tabMood-style recommendation tabCurious Netflix users
DoorDash / Uber agent ordersOnly hundreds of agent orders vs millions of daily tripsReality check for “agent books my life” demos
Google Project SuncatcherFirst AI response from a TPU in orbit (prototype)Interesting, not actionable yet

Why does Matt say this week felt slower?

Matt’s theory: US elections, a soft patch in AI stocks, and Anthropic’s approaching IPO have slowed big launches. Labs may also be waiting to see rivals move. Treat that as commentary, not a forecast.

What should a small business do with these updates?

  1. Use one free or already-paid feature — Google Vids’ three free clips, or Images 2.6 if you already have ChatGPT.
  2. Skip multi-hundred-GB downloads unless you have the VRAM and a clear fine-tuning goal.
  3. Keep agent phone calls in the lab until you have a script, a disclosure and a human fallback.
  4. Do not connect a bank account to Claude until you have read the permissions and decided the upside beats the risk.
  5. Bookmark, don’t rebuild — wait for official notes before changing workflows for cyber models or huge open weights.

What does the video not confirm?

🔎 Unconfirmed / rumour: Hardware requirements, Hugging Face demo performance, DoorDash/Uber earnings figures, Project Suncatcher details and any “imminent IPO” timing are as stated in the video. We have not verified them. Optimizely Virtual Teammates appears as a sponsored segment.

Watch: the source video

What this video covers: Matt Wolfe’s weekly AI news: ChatGPT Images 2.6 GIFs, Pruna PH3 Image, Gemini Omni 1.2 Flash and free Google Vids clips, Pika H3 Flex, Meta Muse voice calls, Grok Bot phone calls, plus a rapid-fire list from GPT-6 Cyber to Claude bank connect and orbital TPUs.

Video: Matt Wolfe (@mreflow). The Optimizely Virtual Teammates segment is sponsored. This summary and the opinions in it are our own. Speeds, prices, hardware figures and usage claims are as stated in the video and have not been independently confirmed.

Frequently asked questions

Is ChatGPT Images 2.6 free?

It is available in ChatGPT and the API according to the video. What you can generate depends on your ChatGPT plan and remaining usage.

How many free AI videos can I make in Google Vids?

Matt says any Google account can generate up to three 15-second Gemini Omni 1.2 Flash videos per month in Vids.

Should I download Pika H3 Flex or PH3 Image?

Only if you have the hardware (or a cloud GPU budget) and a reason to fine-tune. Otherwise use the Hugging Face demos.

Are AI agent food and ride bookings common yet?

Not according to the DoorDash and Uber figures Matt cites. Treat those demos as early.

Did GPT-6 Cyber launch for everyone?

No. The video says access is limited to cyber-security professionals.

Is connecting a bank to Claude safe?

Anthropic describes it as read-only via Plaid and says it does not train on the data. You still share sensitive information — decide carefully.

The bottom line

This was a week of useful edges, not a single must-buy product. ChatGPT Images 2.6 and free Google Vids clips are the easiest wins. Open-weight models and agent phone calls are interesting, but they need hardware or process before they help a small business. Pick one update, test it once, and keep your stack portable.

author avatar
Garry Knight
I'm Garry Knight, the person behind Prodify Digital. I write about email list building, email marketing, SEO, AI search and the tools that connect them. My aim is to make online marketing easier to understand, so creators and small business owners can make informed decisions about building an audience and keeping people engaged. Here you'll find straightforward guides and product reviews that explain what something does, where it fits and which limitations matter. The focus is on clear explanations and useful next steps—not hype, shortcuts or promises of easy earnings.

Leave a Reply