Forget the model race for a second. The news you can use this week fits in a loop: ChatGPT now makes GIFs, remembers your product photo, and does both faster than before. Underneath it, two big companies spent a combined few billion dollars so that an AI can answer the phone. Here's the one you can't ignore, then the rest.
ChatGPT Images 2.6 makes GIFs and remembers your product
On November 4, OpenAI shipped Images 2.6 inside ChatGPT. Three things changed. You can turn a prompt or an uploaded photo into an animated GIF. You can save reference images of a person, a pet or a character and reuse them in later creations without uploading again. And OpenAI says generation is about 50% faster than the previous model. It is available on every ChatGPT tier, on web, desktop and mobile. Developers get two API models, Flare and Sunburst, with Flare producing a low-quality image in about six seconds.
Why does a GIF matter to a shop that sells scarves in the DM? Because a GIF is the smallest unit of motion. A product that turns, a lid that opens, a bag that swings. You drop it in a story, pin it in a broadcast channel, or send it in a DM when someone asks "what does it look like from the side". No video editor, no render queue, a few seconds of wait.
The saved references are the quieter win. Sellers have been fighting the same problem for a year: the AI redraws your product a little differently every time. A saved reference does not fix that completely, but it stops you from re-uploading and re-explaining. Keep your five best product shots as references and your output stops drifting between posts.
One caution. OpenAI has not published frame or length limits for GIFs made inside ChatGPT, and the faster API models cost more per image at low and medium quality than the standard ones, according to early pricing write-ups. For your Instagram feed, use the ChatGPT app and ignore the API prices.
Salesforce buys Cresta, Five9 buys Level AI: the phone is next
On November 2, Salesforce signed an agreement to acquire Cresta, a voice AI company whose agents handle customer calls from start to finish. Salesforce did not disclose the price. The Information reported about $2.7 billion. It is Salesforce's third customer-service deal in two weeks, after Fin for a reported $3.6 billion. Then on November 5, Five9 said it will buy Level AI for about $325 million in cash. Level AI handles more than a billion customer interactions a year for companies like Gusto and VistaPrint.
Read the pattern, not the prices. Two years ago, "AI customer service" meant a chat widget. In 2026 the big vendors are paying billions for voice: an AI that picks up the call, understands the problem and finishes it. Text was the first thing AI got good at. Phone is the second.
For an online seller this is good news with a delay. These are enterprise products built for Comcast and Marriott, not for a shop with 30 orders a day. What reaches you first is the cheaper version of the same idea: voice messages. Your customers already send them in the DM. If your assistant can hear a voice note, read the order out of it and reply, you have 80% of what Salesforce just paid for. We wrote about this in how AI listens to voice messages and takes the order.
The other lesson is in what Cresta also sells: real-time coaching for the human on the call. The AI does not replace the person; it whispers. For a DM shop that means the assistant drafts and the admin approves on the hard conversations. Know when the AI should hand off to a human and you get the best of both.
Google Vids gets 15-second clips, on the cheapest Workspace plan
On November 5, Google put Gemini Omni 1.2 Flash into Google Vids. The slider now runs from 3 to 15 seconds, up from the 10-second clips before, with better quality, steadier characters and more natural voices. The rollout is full, 1 to 3 days, and the list of who gets it is long: Business Starter, Standard and Plus, every Enterprise tier, and personal accounts on Google AI Pro and Ultra.
Fifteen seconds is not a random number. It is a story. It is a short reel. It is the length of "here's the product, here's the price, here's the code". The old 10-second cap forced you to stitch; the new one lets a single generation carry a product post on its own.
And the Business Starter line is the part I'd underline. Google's AI video is usually gated behind the expensive plans. This one is on the entry plan too. If you already pay for a Workspace mailbox, you now have a video generator in the same subscription. Try it before you buy a separate video tool. For what to put in those 15 seconds, our piece on turning story reactions into sales covers the follow-up that actually makes the clip pay.
Quick takes
- Gemini reads your Google Forms. From November 2 (Rapid Release) and November 19 (Scheduled Release), Gemini in the Forms Responses tab finds trends and suggests next steps, and it runs code on the data to do it. If you collect orders or feedback in a Form, that is a free analyst. Not on Business Starter, though.
- Goldman Sachs pulled AI coding tools from traders and bankers. Bloomberg reports the bank now limits Claude Code and GitHub Copilot to its roughly 12,000 engineers after a productivity and risk review. The lesson for a small shop is the same one at a different scale: give the AI the jobs it is clearly good at, measure, and do not hand it everything on day one.
- Meta's business terms are still set for November 20. Nothing new this week, which is the point. The clause that lets Meta use content from websites tied to business accounts has not been removed, and Meta has said the date could slip again. If you run a shop on Instagram, this is the one to re-read before the 20th.
The week's theme is motion and voice: an image that moves, a video that fits a story, an AI that picks up the phone. The part that sells for you is still the same one, the reply in the DM. If you want that reply handled by an assistant that works from your own catalog in Instagram, WhatsApp and Telegram, see how Vardast does it.