Google released Nano Banana 2.1 on 6 October, and through the API a 2K image now costs about half what Nano Banana 2 did. It is rolling out in Flow and Google Ads too, so the same model now sits in the tool your team concepts in and the tool your media buyer launches from. The price drop matters less than the consistency claim: four characters and ten objects held across one job.
Nano Banana 2.1: half the price, more references, same weak spot
Google’s model card, published 6 October, lists the Gemini app, AI Studio, the Gemini API, AI Mode in Search, Google Ads, Flow and Stitch as surfaces. It takes up to 14 reference images and keeps up to 4 characters and 10 objects consistent (Google DeepMind, 6 Oct). On the API price list a 1K image is $0.0336 and a 2K image $0.0504, against $0.067 and $0.101 for Nano Banana 2; 4K goes from $0.151 to $0.113 (Gemini API pricing). The same card names the limits: “poor text rendering in small text (often blurry in 1k model)” and character consistency that “remains imperfect”.
My read: For the Shirin Asal Christmas greeting I generated the images with Nano Banana and built each sequence from them in Kling. The image stage is where a character either holds or drifts, and drift there costs you every shot after it. The header image on this brief comes from Nano Banana 2 every morning, so I will test 2.1 on the same prompts before switching. Do not trust the small print on a pack: composite the real label in post.
Meta turns catalog photos into product video
At Advertising Week New York on 6 October, Meta said generating video from static assets in Advantage+ creative is now generally available. Catalog advertisers in a beta can turn product images into product video at scale. Ads Creative Studio is rolling out more broadly, with tools to change hooks, audio, clip length, captions and end scenes (Meta for Business, 6 Oct).
My read: This is a long-tail tool, not a hero-film tool. In POFAN BBQ the flavour opens as real fire on a grill before the pack appears, so the taste is set before the logo. An animated packshot cannot make that choice. It can keep 200 SKUs moving that would never get a film. Decide which products get directed work and which get the machine. Then check every generated end scene against your pack, because Meta says nothing about how it handles your label.
Higgsfield Ad Multiplier rebuilds an ad around your product
On 6 October Higgsfield released Ad Multiplier with a visual interface. You upload one reference ad of 4 to 30 seconds and up to 20 images, then swap the product, cast, location, script, hook or CTA while the original structure stays. Output is 480p, 720p or 1080p (Higgsfield changelog, 6 Oct).
My read: Structure is the cheap part of an ad. The idea is the expensive part. Hit Wafers rests on one sign, a finger on the lips, and the film works because the dad takes that sign over. Swap the product and the sign survives. Swap the cast carelessly and it is gone. Use your own past ads as references, not a competitor’s. 1080p is also a ceiling: fine for social cutdowns, not for a broadcast master.
What I’d do today
Pick one product with a hard pack, small print and a character. Run the same three prompts through Nano Banana 2 and 2.1 at 2K, and write down where the label breaks. That one sheet tells you whether to switch, and what to keep compositing by hand. If you want the production side of that, this is how I work.
Sources
- Google DeepMind — Nano Banana 2.1 model card, 6 Oct 2026 — deepmind.google
- Google — Gemini API pricing (Nano Banana 2.1 and Gemini 3.1 Flash Image), checked 7 Oct 2026 — ai.google.dev
- Meta for Business — Advertising Week New York 2026: New AI Capabilities, 6 Oct 2026 — facebook.com/business
- Higgsfield — Changelog: Ad Multiplier, 6 Oct 2026 — higgsfield.ai