Nano Banana 2.1. A Panoramic Still Is Not a Clip

Nano Banana 2.1. A Panoramic Still Is Not a Clip

On 6 October 2026 Google marked Gemini Nano Banana 2.1 generally available in the Gemini API. The model code is gemini-nano-banana-2.1. The card says it is an update to Nano Banana 2 (gemini-3.1-flash-image): image generation and conversational editing, Flash-level speed, 1K / 2K / 4K stills (default 1K), fixed tiling on wide and panoramic ratios, better text and infographic layout, up to 14 reference images, Google Search grounding, and Thinking levels. The same card lists output as image and text. Audio generation is not supported. That is a stills desk. It is not Veo, not Gemini Omni Flash, not a signed script, and not Monday’s upload. ClipCanva is not affiliated with Google, VEED, or Kapwing. Lock the spoken line on the AI script generator. Steal shot language from prompt ideas or the ads and Shorts prompt pack. Compare stills models on the Nano Banana and Nano Banana 2 guides, then animate a locked frame on image to video or generate motion on the AI video generator. Do not hit generate-video because the infographic now survives 8:1.

Product facts (confirmed 10 October 2026)

Source	Public claim	Workflow implication
Gemini API release notes (6 Oct 2026)	GA: gemini-nano-banana-2.1. Update to Nano Banana 2. Improved visual quality, prompt adherence, multi-turn character consistency, text rendering, and wide / panoramic ratios 1:4, 4:1, 1:8, 8:1 across 1K, 2K, and 4K. gemini-3.1-flash-image is deprecated with no shutdown date announced. Migrate new work to 2.1.	A stills endpoint change. Not a video model swap. Do not brief “Nano Banana now makes the clip.”
Gemini Nano Banana 2.1 model card (last updated 2026-10-06 UTC)	Inputs: text, image, video, PDF. Output: image and text. Image generation supported. Audio generation not supported. Search grounding supported. Thinking supported (minimal, medium default, high). Multi-image fusion: up to 14 references — character consistency for up to 4 characters, object fidelity for up to 10 objects. Default still size 1K. Tiling fix is for those panoramic ratios at 2K and 4K. Token limits: 131,072 in / 32,768 out. Efficient counterpart to Gemini 3 Pro Image (Nano Banana Pro).	Video on the input side is not video out. Fourteen references are stills fusion, not a multi-shot timeline.
Image generation (Nano Banana)	Nano Banana is Gemini’s native image generation. 2.1 is the primary high-efficiency workhorse. All generated images include a SynthID watermark. 2 Lite is fastest / cheapest and not optimized for multiple references or multi-turn sequential editing. New projects: use 2.1, not Nano Banana 2. Legacy gemini-2.5-flash-image: migrate toward 2 Lite.	Watermark ≠ YouTube label. Lite is a draft desk, not the 14-reference pack.
Generate images using Imagen	Imagen models are shut down in the Gemini API. Google tells developers to migrate to Nano Banana.	Do not put imagen-* in a 2026 runbook.
Video generation in the Gemini API	Gemini Omni Flash: fast video generation and conversational video editing from text and images. Veo 3.1: video with native audio, extension, frame-specific generation. Separate family from Nano Banana.	If the job is motion or dialogue, you are not on 2.1.
Gemini API pricing	Paid 2.1: input $1.50 per 1M tokens (text / image / video). Output $7.50 per 1M (text and thinking) and $30.00 per 1M (images) — Google’s equivalents $0.0336 / 1K, $0.0504 / 2K, $0.113 / 4K. Free tier listed Not available. Grounding: 5,000 free search requests / month shared across Gemini 3 and newer, then $14 / 1,000.	Price the still. 4K panoramic is not “the same as 1K.” Do not invent a ClipCanva credit map from this table.
VEED AI Image Generator	Own URL. Headline: generate an image, then turn it into a video. Separate AI Video Generator for text, scripts, or images.	Writer / stills / renderer stay different URLs.
Kapwing AI Image Generator	Own URL. Text and reference pictures. Lists models including Google Nano Banana. Separate AI Video Generator for text, images, and scripts.	Kapwing still splits image and video. Listing Nano Banana on an image URL does not make 2.1 a clip desk.

Google did not publish, on the 2.1 card, native audio, clip duration, a first-and-last-frame video interpolator, or a statement that 2.1 replaces Omni Flash or Veo. If a capability is not on the page you are briefing from, it is not a brief.

What 2.1 is — and what it is not

Operators will collapse four objects into “Google shipped Nano Banana 2.1, so the restock video is done.” The public pages refuse that.

Surface	What the public page says it does	What it does not do	Fail condition
Nano Banana 2.1 (6 Oct 2026 GA)	Generate and conversationally edit stills; 1K–4K; panoramic tiling fix at 2K/4K; infographic text; up to 14 refs	Output video or audio; replace Veo / Omni Flash; sign Legal copy	Shipping a 4K PNG as the YouTube master
Nano Banana 2 (gemini-3.1-flash-image)	Previous workhorse. Deprecated, no shutdown date	Stay the default for new projects	Leaving 2 in the runbook because “it still answers”
Nano Banana 2 Lite	Fast / cheap stills. Not optimized for multi-ref or multi-turn sequential edits	Hold four characters across a pack	Using Lite for the 14-image fusion, then blaming 2.1
Nano Banana Pro (Gemini 3 Pro Image)	Premium stills: world knowledge, localization, brand consistency	Become 2.1’s video mode	Paying Pro rates for a shot 2.1 already lost on motion
Gemini Omni Flash / Veo 3.1	Video generation (Veo: native audio)	Repair a warped label on the still	Prompting 2.1 “make it move”
VEED / Kapwing image URLs	Stills first; video on a different URL	Disclose on YouTube or lock your SKU	Hitting “generate video” from an unsigned still
SynthID on every Nano Banana still	Invisible watermark on generated images	File YouTube AI use, or prove a miss is photography	Treating the watermark as the platform label

If you are calling gemini-nano-banana-2.1, you are on a stills desk. If Omni Flash or Veo is in the model string, you are on a motion desk. If the bottle still needs a legal sentence, you are on a writer. If you cannot tell which, do not publish.

Pick the desk before you spend the 4K still

The fail is treating “panoramic tiling is fixed” as “the eight-second restock is covered.” Map the job.

Job	Where it actually runs	What you must supply	What the system may invent	Fail condition
One claim the clip must carry	AI script generator — or VEED / Kapwing if that is already your writer	Audience, seconds, the signed sentence	A second CTA, a health claim, a competitor name	Skipping the card because “the infographic has the words now”
Recap of a long cut	AI video summarizer	Source cut, claims to keep, claims to strike	Compression that changes meaning	Feeding the webinar into 2.1 as if video-in meant video-out
Still that must stay on-model	2.1 in the Gemini API, or the labeled generator on Nano Banana 2 / Nano Banana	Approved pack shot or up to 14 refs; crop; 1K vs 2K vs 4K	Extra SKU, warped type, a fifth character	Using 2 Lite for a four-character pack
Motion from a locked still	Image to video	The signed still as frame one	Camera path, extra hands, a spoken line you did not approve	Animating take three because 2.1 made a prettier 8:1
New motion without a locked frame	AI video generator or a named video model on Compare	Locked caption, crop, seconds	Props, voices, a second location	Briefing 2.1 as the compare winner for video
Model pick	Compare	Job = still vs clip vs native audio	A leaderboard screenshot	Picking 2.1 because the changelog said “GA”

If the brief is “fix the banner tiling at 8:1,” that is 2.1. If the brief is “the two-pack restocks Friday on YouTube,” that is script, still, then image-to-video. If the brief is “Nano Banana 2.1 will handle the clip,” you do not have a brief.

VEED still keeps AI Image Generator as its own URL, then offers turn-it-into-a-video after the still. Kapwing still keeps Image Generator and AI Video Generator as different URLs, even while the image page names Google Nano Banana among models. Neither page says 2.1 outputs video. ClipCanva’s Nano Banana guide still documents the older Gemini 2.5 Flash Image identity and labels the embedded generator Nano Banana 2 — do not brief that widget as 2.1 unless the live picker says gemini-nano-banana-2.1.

Creator / operator checklist

1. Read the 6 October line out loud. GA. gemini-nano-banana-2.1. Stills. Image and text out. Audio not supported. If the next sentence in Slack is “we can drop Veo,” stop.
2. Name the string. 2.1 ≠ 2 ≠ 2 Lite ≠ Pro ≠ Omni Flash ≠ Veo 3.1. Screenshot the model code, not the nickname.
3. Cap the spoken line before you generate a still. Use the AI script generator. One sentence. Strike “#1,” ingredient claims, and competitor names you cannot defend. 2.1 can render the words on the image. It will not speak them. Omni / Veo might, and they will say whatever the prompt implied, including the lie.
4. Pick 1K, 2K, or 4K on purpose. Default is 1K. Google’s paid equivalents are $0.0336 / 1K, $0.0504 / 2K, $0.113 / 4K. Do not 4K a draft crop.
5. Use panoramic ratios for banners, not for a 9:16 master. The tiling fix is 1:4, 4:1, 1:8, 8:1 at 2K and 4K. A YouTube Short is still a different crop. Generate the hero still in the crop you will animate.
6. Count the references. Up to 14 images; 4 characters; 10 objects. 2 Lite is not the tool for that pack. Extra faces past four are not “the model will figure it out.”
7. Grounding is a search bill, not a legal review. Search grounding is supported. Paid: 5,000 free searches / month shared, then $14 / 1,000. A grounded infographic is not a citation Legal signed.
8. SynthID is on the still. The image-generation guide says every generated image includes a SynthID watermark. That is not YouTube Studio AI use. Detector vs YouTube disclosure is a different desk.
9. Imagen is shut down in this API. If the runbook still says imagen, rewrite it before the next campaign, not during it.
10. Bill the right object. A script, a 2.1 still, an image-to-video pass, and a YouTube disclosure are four things. Do not hide all four under “we upgraded Nano Banana.”

Worked example: a Friday restock that does not start in 4K panoramic

Say the job is a YouTube restock cutdown. The two-pack ships Friday. Legal line approved: Same bottle. Same seal. Restock Friday. Someone read the 6 October changelog, saw 8:1, saw 4K, saw 14 references, saw “GA,” and told the channel “Nano Banana 2.1 is live, so the video is covered.”

That sentence mixes five desks.

The spoken line is a writer object. Put the approved sentence on the AI script generator. 2.1 can typeset those words on a pack shot. It cannot say them. If you already live in VEED or Kapwing, their script tools are still script tools. Copy the signed line. Do not hit generate-video from the unsigned still.

The beat list is a card object. Steal shot language from prompt ideas or the ads and Shorts prompt pack, then write four lines, not a chat novel:

• 0s — kitchen, bottle already on the board (do not invent a second SKU)
• 2s — hand in, label readable
• 4s — pour or use, same bottle
• 6s — end card with the signed sentence only

The still is a frame object. If you need a new pack shot, call 2.1 (or the labeled stills generator you actually have). Crop 9:16 if the upload is a Short — not 8:1 because tiling is fixed. Default 1K until Legal signs the label. If you must fuse a bottle, a hand, and a kitchen, stay inside 14 refs / 4 characters / 10 objects. Do not use 2 Lite for that pack. Confirm the live picker: ClipCanva’s Nano Banana pages currently label Nano Banana 2, not 2.1.

The motion pass is a clip object. Feed the signed still to image to video. If you still need picture-plus-motion in one hop, that is the AI video generator or a video model on Compare — Omni Flash or Veo 3.1 on Google’s side, not gemini-nano-banana-2.1. Native audio is a Veo claim on the video overview, not a 2.1 claim.

The watermark is a check object. 2.1 stills carry SynthID. That does not file YouTube AI use. A photorealistic restock in a real kitchen is still a disclosure job. Do not skip Studio because the PNG is 4K.

FAQ

Does Nano Banana 2.1 generate video now? No. The 6 October GA notes and the model card describe image generation and conversational editing. Output is image and text. Video is listed as an input. Motion is Gemini Omni Flash or Veo 3.1.

Can I leave production on Nano Banana 2? Google says 2 (gemini-3.1-flash-image) is deprecated with no shutdown date announced, and that new projects should use 2.1. Existing calls may still answer. That is not a promise they will keep answering. Screenshot the string you actually ship.

Is 2 Lite the cheap way to do the same 14-reference pack? No. The image-generation guide says 2 Lite is not optimized for multiple reference inputs or multi-turn sequential editing. Cheap drafts, then 2.1 (or Pro) for the pack you will animate.

Does a SynthID watermark on the still mean YouTube is covered? No. The image-generation guide says generated images include SynthID. YouTube Help is a separate Studio disclosure for realistic generated scenes. Run both when the upload is photorealistic GenAI.

Will ClipCanva’s Nano Banana page call 2.1 today? Not as of this writing. Nano Banana documents Gemini 2.5 Flash Image and labels the embedded generator Nano Banana 2. Nano Banana 2 documents Gemini 3.1 Flash Image. Treat the live picker as source of truth. The stills-then-image to video desk does not depend on the nickname.

Do not mix these objects

Object	Tool	Done when
Signed line	AI script generator	One sentence Legal will defend
Shot card	Prompt ideas	Beats with times, no extra SKU
Locked still	gemini-nano-banana-2.1 or the labeled stills generator	Label readable, crop matches the upload, refs inside 14 / 4 / 10
Motion	Image to video or AI video generator	One scene, approved audio
Model pick	Compare	Still vs clip vs native audio named
Platform disclosure	YouTube Studio AI use	Realistic GenAI labeled; watermark not used as the form

Google made the stills workhorse 2.1. It did not make the clip desk 2.1. Write the line, lock the frame in the crop you will ship, then animate that frame — not the panoramic banner you generated because tiling finally held.