Google Gemini 3.0 spotted in Vids – Nano Banana Pro powered by 3 Pro

A hidden panel in Google Vids labels “Nano Banana 2” as “Nano Banana Pro” and states it’s powered by Gemini 3 Pro, with a “Try it” CTA and copy promising “richer, more dynamic video images.” This points to image generation landing directly inside Google’s video editor—useful for storyboards, inserts, and b‑roll without leaving the timeline.

For creatives, this hints at a one‑roof workflow: Veo for shots, Nano Banana Pro for stills/visuals, and Vids as the canvas. Naming also aligns the consumer-friendly “Nano Banana” brand with a concrete model tier (Gemini 3 Pro).

A brief on-device clip displaying “Gemini 3.0” is circulating, following up on mobile Canvas early sightings; several Google‑adjacent voices say “next week” would be the moment, so teams are bracing for changes to image/video tools and on‑device assistants. This is unofficial, but the cadence of hints has tightened.

If you plan launches around Google tooling, line up quick tests and prompts so you can benchmark outputs on day one. The social chatter framing it as imminent matters because asset pipelines and prompts can be prepped now.

A community mini‑app built on Nano Banana lets you upload a photo, choose or write a style, tweak options, and then it keeps producing images in that style until you halt the run. This is handy for batch campaigns, lookbooks, or character sheets where you want consistent variants without babysitting the loop.

You can peek the shared Gemini link to see how the flow is wired and adapt it for your own presets and client palettes. Gemini share

A designer post suggests Gemini 3 can emit high‑quality SVGs, hinting at native vector output for icons, logos, and UI components—key for clean scaling and editability in Figma/Illustrator workflows. It’s unconfirmed, but if true it cuts raster‑to‑vector cleanup loops for brand kits and mobile UI packs.

If you handle brand systems, prepare a small SVG test battery (logo variants, monochrome icons, stroke‑based glyphs) to validate path hygiene and grouping the moment access lands.

Creators are running dry‑runs for Gemini 3 prompts while showcasing Nano Banana’s current range—one example turns a kid’s photo into a convincing LEGO‑style minifig product shot, useful as a sanity check for style fidelity and studio‑lighting control before the next model lands.

Use the prompt‑call threads to harvest phrasing that controls materials, shadows, and lens feel, then save a handful of A/B scripts for immediate side‑by‑side tests once Gemini 3 opens up.

InVideo’s FlexFX takes a single photo and spits out on‑trend motion effects in about 60 seconds, no After Effects or CapCut required How to thread, with a short clip showing the photo‑to‑VFX workflow in action Workflow demo.

The recipe is dead simple: in “Agents & Models,” pick “Quantum Leap” or “Drip Hop,” upload a photo, and generate—creators say the results look like they hired a $2K freelancer, but it’s all inside InVideo How to thread No editor needed. For social teams, the pitch is clear: built for virality and speed, with promises your feed will look different in a week Virality pitch. Try it now from the product page InVideo AI page.

Higgsfield’s Black Friday promo hits 65% off and advertises “unlimited image models” access for 365 days, with a 9‑hour 250‑credit social boost for follow/retweet/comment.

For performance marketers, the math is simple: lock in a lower seat now if you rely on rapid ad variants and face/character swaps across product lines.

Higgsfield’s Click‑to‑Ad flow is making the rounds again: drop a product link, the tool pulls assets, you choose a style, and it outputs a UGC‑style promo. This follows product-to-ad, which highlighted the initial URL‑to‑video builder.

The product page lists 25+ ad presets plus AI avatars and caption styles for platform‑native looks AI ad generator. For affiliates and small stores, it’s a practical way to scale creative variations without a full edit pass.

Leonardo is pushing a new Product in Scene Blueprint that auto-places your item into a context-appropriate, styled photo for ad use. The pitch: upload a product, get a pro-looking placement with copy-ready framing in seconds.

For solo marketers and small shops, this cuts the scramble for props, location, and retouching. It’s a fast way to A/B test backdrops and mood before spending on a real shoot.

Leonardo also teased a Product Spin Video Blueprint that generates sleek 360‑style rotations from a single image. That’s the classic e‑com spinner without the multi-angle shoot.

If your catalog relies on motion thumbnails or PDP loops, this is a cheap way to standardize spins across SKUs and speed up seasonal refreshes.

Pictory’s Black Friday deal is live: 50% off annual plans and 2,400 AI credits, with access hooks into ElevenLabs and Getty assets. It’s positioned as “pay for 6 months, get 6 free.”

If you’re sitting on blogs, webinars, or decks, this is a lower‑cost window to test the repurpose‑to‑video pipeline and standardize brand styles across a year Pricing page.

A creator cut a full music video on OpenArt using Kling 2.5 Turbo with Start–End Frame, reporting cleaner consistency and more controlled motion Creator verdict, following up on Start–End frames improving continuity. For filmmakers, this reads as less flicker and steadier identity across shots—exactly what hurts long edits.

A new ‘Video Prompt Benchmark’ montage landed, stacking text prompts against their generated visuals so you can eyeball strengths and misses fast Teaser clip. For teams tuning prompts and comparing models, this kind of side‑by‑side is a quick sanity check before deeper tests.

A kinetic ComfyUI workflow auto‑sequences images to music beats for rhythm‑locked motion pieces Project note. The live demo shows a pulsing visualizer synchronized to the track—useful as a template node graph for editors needing music‑tight loops.

A shared ALT prompt for Grok Imagine outputs a five‑panel ‘cinematic fashion’ sequence with coherent styling and subject continuity across shots Prompt + grid. It’s a handy base for moodboards and storyboard beats before animating.

Azed_ai’s reusable prompt—“A stylized sculpture of a [subject], entirely constructed from folded metallic paper…”—is yielding crisp, faceted forms across subjects (violin, face, owl, skull), making it a versatile kit for product‑look explorations and poster art Prompt template. Community riffs extend it to sax, turtle, spaceship, and rose while keeping specular discipline; a third‑party check scored one example a 90% prompt‑to‑image match Prompt score card.

A new Midjourney V7 style reference (--sref 1575542916) reliably produces a classic Studio Ghibli–era anime look with hand‑painted backgrounds and a steampunk adventure vibe, ready to drop into your prompts today Style ref details. This is useful when you need consistent cinematic anime frames without post‑tweaking.

Another Midjourney V7 style ref arrives: --sref 651772730 with --chaos 22, --ar 3:4, --sw 500, --stylize 500 yields warm, desaturated watercolor collages with strong orange accents and soft edges—handy for storyboards and mood posters Prompt settings. This builds on a wave of V7 style refs, following up on neo‑retro anime which mapped a techno‑gothic anime look.

Bri_guy_ai shares a detailed portrait prompt for an 80s retrofuturism look—puffed sleeves, multi‑hue palette, soft vintage photo feel, and a playful cat print—targeted at an 85:128 aspect ratio for tall editorial frames Full prompt. It’s a fast way to standardize a cohesive, nostalgic portrait set.

A Munich court ruled OpenAI violated German copyright by training ChatGPT on lyrics from nine songs, raising the stakes for dataset provenance in the EU Court ruling report. Following up on NYT demand request for 20M chats, this adds concrete liability risk and signals stricter scrutiny of training data disclosure and rights-holder permissions for models deployed in Europe.

Disney is preparing AI creation tools inside Disney+, but only for content using Disney-owned characters and worlds—no original IP, no crossovers, and strict platform control UGC limits post, with details echoed again later Follow-up link. A creator write‑up frames this as a corporate sandbox and flags immediate pushback from subscribers questioning fan‑made content sitting next to pro productions Studio blog post. For creatives, expect tight moderation, unclear export rights, and engagement locked to Disney’s ecosystem rather than open storytelling.

A quick showcase praises Grok Imagine’s pace of updates; worth tracking if you’re betting on frequent style and motion improvements Speed praise.

Fast motion, bold color wipes, and clean character design make this anime micro‑reel pop for action intros and teasers Anime demo.

A boy‑and‑robot vignette cycles through multiple anime looks while staying coherent, useful for exploring style tests on a fixed story beat Iron Giant homage.

A quick visual joke begs for animated profile pics—good template for looping, punchy identity clips on X/TikTok Animated pfp gag.

A minimal fabric shot shows tasteful motion and lighting—great as a reusable motif for titles, brand bumpers, or mood reels Cloth physics demo.

Short, musical loops suggest where integrated audio could go for micro‑stories; creator hopes for Suno‑like song generation next Stop‑motion loop.

Tender framing and gentle motion sell a kiss-in-flowers moment; handy reference for mood boards and lyric videos Romance clip.

Simple sky pass with a soaring eagle—useful as a background plate or breathing space between heavier scenes Eagle fly‑by.

A new how-to shows Adobe Firefly’s Text to Sound Effects turning prompts into bespoke SFX fast—useful for punch-ins, transitions, and ambience without stock hunting Tutorial post. For small teams and solo creators, this trims timeline cleanup and keeps your sound consistent scene to scene.

Pictory published a step‑by‑step Text‑to‑Speech guide that auto‑syncs narration to scenes—good for explainer voiceovers and quick story cuts TTS tutorial. It’s positioned as “no mic needed,” with language/voice options and timing controls to fit your edit Academy post.

A short demo reimagines Korn’s track in flamenco style, with a full workflow and prompts promised later—useful for musicians testing genre‑transfer cover pipelines Cover demo, and you can hear the longer version on YouTube YouTube track.

A Firefly Speech demo focuses on rapid character voicing and light dubbing—drop in visuals, script, and generate in-voice lines for reels or short films Feature demo. This is handy for temp tracks and indie dubs before bringing in actors.

OpenArt shared winner reels from NYU, MIT, and Harvard and issued a last‑day call for entries to its Music Video Awards (deadline Nov 16). The program centers on AI‑made music videos around an "Emotions" theme, with artist‑picked winners Final day call, and rules posted on the awards page Awards page.

• NYU: “Grandma Knows Best” by @IrinaFachin NYU winner
• MIT: “Alone Again” by @reubenuntuned MIT winner
• Harvard: “Alone Again” by a creative team Harvard winner

Why it matters: these reels are strong reference points for pacing, coherence, and prompt design under contest constraints—use them to calibrate your own submissions or as teaching examples for student teams.

Filmmaker PJ Accetturo is attempting a commercial shoot mid‑flight—15,000 feet up—with 15 hours to deliver a Qatar Airways spot, leveraging Starlink connectivity and collaborating with creators RPNickson and Torey, plus remote support from Genre Shoot announcement.

Why it matters: real‑time, in‑air production tests whether AI‑assisted editing, review, and asset swaps hold up under tight bandwidth and hard deadlines—useful signal for travel, sports, and event workflows that can’t wait for landing.

Krea is staging an in‑person AI art gallery in Barcelona on Tuesday, Nov 18, with RSVP required and a casual art‑and‑drinks format. Following up on Infra talks community meetup, this signals a steady cadence of creator events.

Sign‑up details and guest list are live via Partiful Event invite, with the page outlining the format and access rules RSVP page.

Why it matters: local gallery nights are great for scouting collaborators, sharing reels in person, and getting unfiltered feedback before publishing more widely.

Elon Musk outlined a Grok 5 roadmap aiming for ~6 trillion parameters, multimodal with real‑time video, and a Q1 2026 launch window. For filmmakers and designers, the headline is potential higher‑fidelity, frame‑coherent reasoning on long clips—if training and inference costs land where creatives can actually use it.

Interview recap

Google Research sketched a continual‑learning paradigm that layers fast and slow learning processes, introducing a HOPE model that adapts during inference while retaining old skills. Reported results include 15.11 WikiText confusion and 73.29% on commonsense questions, alongside reduced catastrophic forgetting—useful for tools that must keep style and character consistency over long projects.

Blog announcement

Meta’s Reinforcement Learning with Verifiable Rewards (RLVR) highlights that LLMs learn differently under verifiable reward signals than with standard RLHF, suggesting a path to more reliable multi‑step reasoning and tool use. For creative pipelines, that could mean steadier shot planning, prompt chains, or editing agents that degrade less over long runs.
Paper highlight

Anecdotal reports point to mathematics as an early field converging on LLM‑assisted research, with claims that GPT‑5 can prove certain lemmas in ~20 seconds versus about an hour for a human and even tackle an open convex optimization problem. If this sticks, it shortens proof exploration loops for technical creators who depend on formal methods, graphics math, or procedural animation.
Professor roundup