Changelog

v1.59.1 · 2026-10-10 — Tools are "use and go" 🧹

• 🧹 Tool results no longer publish to the gallery: the gallery is a prompt community, so output from Cover Maker, Relight and Film Emulator stays out of it; you can still download, save images to your generations, use them as a reference, send them to the canvas or hand them to another tool
• 🎬 Video results are download-only: MP4 / WebM exported from Film Emulator are no longer saved to your generations; image saving is unchanged
• 🛡️ Safer submissions: gallery submissions now have quotas (at most 5 pending at once, 10 per day), and a bug where deleting a generation could affect a shared file has been fixed

v1.59.0 · 2026-10-10 — New tool: Film Emulator 🎞️

• 🎞️ Film Emulator at /tools/film-emulator: give photos and short videos a film look — color, grain, halation, bloom and vignette, plus exposure flicker, dust and gate weave for video; switch between original / film / split-view and tweak while the video plays
• 🎨 40 styles in three groups: 18 one-click grading looks (teal & orange, tobacco, bleach bypass, day-for-night, seventies, telecine…), 12 film stocks modeled on published manufacturer tone curves (Daylight 50D, Tungsten 500T, Portrait 400, Slide 50, Mono 400 and more — no brand names, no commercial LUTs) and 10 original film bases; every thumbnail is rendered live from your own media
• 🎬 "Use video" straight from the gallery: video cards, the detail view and your own generations now list Film Emulator under "Use video"; image posts reach it from "Use image" too
• 📤 Export MP4 / WebM or PNG: three quality tiers (1280 / 1920 / original), video is re-encoded frame by frame with audio preserved and frame-level progress; sign in to export — free accounts included; videos are download-only (never uploaded or saved to your generations), while processed photos can still be saved
• 🔒 Runs entirely in your browser: nothing is uploaded, no models are called; video export needs a WebCodecs browser (Chrome / Edge), photos work everywhere
• 📌 Sticky tool header: on every tool page the tool name and Export button now collapse into one bar that stays under the navigation as you scroll
• ▶️ Video autoplays when ready: the sample or your own clip starts muted automatically, so you see the look without an extra click

v1.58.0 · 2026-10-09 — New tool: UI Theme Studio 🎨

• 🎨 UI Theme Studio at /tools/theme-studio: nothing to upload — open it and design a UI theme in your browser. Start from one of 5 styles (Soft, Precise, Paper, Airy, Playful), pick or drag on the character grid between airy / dense and calm / bold, then tune color, neutrals, fonts and motion
• 🧩 38 components, real page previews: buttons, inputs, selects, tables, toasts, steps, side sheets and more update live with your theme; switch between the style specimen, a settings page, a data page and the full component library, in light or dark, desktop or phone
• 🤖 Hand off to your agent: download a ZIP — React components (Base UI), the style's open-source fonts, light and dark semantic tokens, design guidelines, page examples and a local check script — then copy a short instruction for Cursor, Claude Code or any coding agent to build from
• 🔒 Runs entirely in your browser: no models, no uploads; your edits are saved as a local draft for 7 days; sign in to download the package
• ✨ An original specimen for each of the 5 styles: Soft is a creator's release planner, Precise an asset library with licenses, Paper a magazine newsroom, Airy a portfolio gallery and Playful a watercolor class; the settings, data and component-library pages switch their titles and sample data with the style too
• 🔤 Open-source fonts: each style pairs two open-source typefaces (Instrument Sans, Geist, Fraunces, Manrope, Nunito and more), shipped in the package with their license
• 🎬 Feels alive: switching styles plays a short transition and light / dark fades smoothly; buttons go press → loading → done, and switches, radios and number steppers all animate — or pick "Still" motion; hover the character grid to preview a cell, click to keep it; upload an image to pull out five brand colors
• 📝 Good to know: it is a design-system starting point for coding agents, not a full design tool — once it is in your project, you and your agent still need to check interactions, layout and real content

v1.55.0 · 2026-10-08 — New model: Nano Banana 2.1 🍌

• 🆕 Nano Banana 2.1 launches: Google's newest Nano Banana image model — sharper in-image text and more consistent characters across images; supports text-to-image and image-to-image (up to 10 reference images), with output up to 4K
• 💳 Pricing: 1K / 2K / 4K = 5 / 8 / 12 credits, same price for text-to-image and image-to-image, no surcharge for aspect ratio or reference count; listed under the "Featured" group in the model dropdown
• 📐 14 aspect ratios + auto: 1:1 / 2:3 / 3:2 / 3:4 / 4:3 / 4:5 / 5:4 / 9:16 / 16:9 / 21:9 / 1:4 / 4:1 / 1:8 / 8:1, plus "auto"; prompts up to 20000 characters
• 🍌 Alongside Nano Banana 2: the existing Nano Banana 2 and Nano Banana 2 Lite stay as they are — switch freely by need and budget

v1.54.0 · 2026-10-08 — New tool: Relight Studio 💡

• 💡 Relight Studio at /tools/relight: upload a photo and drag virtual lights right on the picture to see light fall on faces, walls and objects in real time; three rings around a selected light set brightness, color temperature and range
• 🎛️ 19 lighting presets: soft light, sunset, neon, Rembrandt, butterfly, split, rim light, duotone, dappled leaves, afternoon blinds, window panes and more, each with a live thumbnail — then fine-tune ambience, exposure, shadows and background depth
• 🌿 12 line-art stickers: fern, lotus, koi, moon & stars, a one-line mountain and more; stickers follow the surface and pick up the light — want your own? Make them with Skills on the canvas
• 🔒 Runs on your device: depth analysis and rendering happen in your browser and your photo is never uploaded; the first analysis downloads the depth model once (roughly 70–85 MB depending on your device) and reuses it afterwards — on phones we ask before downloading
• 📤 Export: sign in to export PNG / JPG / WebP, up to 9 MP on a computer and 4 MP on a phone; this is 2.5D lighting from single-image depth, not generative relighting, so hair, glass and other see-through edges may need a manual touch

v1.52.0 · 2026-10 (upcoming) — New tool: Cover Maker + Quick Tools 🧰

• 🆕 Quick tools at /tools: a new tools area, reachable from the Studio dock and sidebar — hover on desktop to see every tool; the four prompt tools now live here too
• 🖼️ Cover Maker: upload, drag in or paste a photo, pick one of six layouts (Mist, Paper, Dusk, Fade, Sidebar, Clear) and the text sits on a progressive blur that keeps it readable; 22 sizes for X, Instagram, Facebook, YouTube, TikTok, LinkedIn and Pinterest, plus 8 filters and effects such as liquid glass, fluted glass, pixelate, ASCII and color halftone
• 🔒 Runs entirely in your browser: layout and export never upload your photo; export PNG / JPG / WebP, or 2× on desktop — the image is only uploaded when you save it to your generations or publish it
• ✍️ AI copy and AI base images (signed in): draft a title, subtitle and small print in one click (included in Pro, daily quota on Starter, a few credits on the free plan), or generate a base image with AI, billed by the model you pick and refunded automatically on failure
• 🔁 "Use image" menu: the old "Use as reference" button on gallery details and generation cards becomes "Use image" — use it as an image or video reference, or open it straight in a tool; when the image is someone else's gallery post, we remind you to check the rights first
• 📤 Keep going with the result: download, save to generations, publish to the gallery, use as reference, send to canvas or compare before / after — all from one action bar

v1.45.0 · 2026-09-23 — New model: Qwen Image 2.1 🖼️

• 🆕 Qwen Image 2.1 launches: Alibaba's new-generation image model — accurate in-image text rendering, understands prompts in any language, and shines on posters, covers and title-heavy layouts; supports text-to-image and image-to-image (up to 10 reference images, referenced in order as "the first image", "the second image")
• 💳 Pricing: 1K / 2K = 6 / 12 credits, same price for text-to-image and image-to-image, no surcharge for aspect ratio or reference count; listed under the "Featured" group in the model dropdown
• 📐 9 aspect ratios: 1:1 / 4:3 / 3:4 / 3:2 / 2:3 / 16:9 / 9:16 / 21:9 / 9:21, plus "auto" in image-to-image to follow the first reference image; prompts up to 5000 characters

v1.44.0 · 2026-09-22 — New canvas tool: Vectorize + new Skill: Capsule-Eye Bot Avatar 🤖

• 🆕 Vectorize (0 credits): select any single image, hit "+" → "Edit" → "Vectorize", and the whole image is traced to SVG in your browser — colour / black & white / grayscale modes, 1024 or 2048 trace size, three detail presets plus an expandable advanced panel; black & white can skip pre-processing to keep fine lines, and non-colour modes can be inverted in one click
• 🖼️ Result on the canvas, SVG always downloadable: the result lands as a transparent PNG node with a connector, and the node's download button hands you the SVG; after a page refresh it re-traces with the original settings on demand. Exported SVGs drop straight into Figma or Illustrator
• 🆕 Capsule-Eye Bot Avatar Skill: new under the Skills panel's Illustration category — upload a portrait or character photo and it's rebuilt as a square minimal icon with two solid black capsule eyes, a mouthless round face and a lower-left corner-peek pose; black (default), matched-colour or white background, mirrorable entrance; defaults to GPT Image 2.5 at exactly the same price as a regular generation
• 🛠️ Agent fix: when a text node produced by "image to prompt" or "write text" is selected together with an image, the Agent now resolves @text1 correctly instead of answering that no text was found

v1.43.1 · 2026-09-16 — New canvas tool: Storyboard sheet 🎞️

• 🆕 One-click storyboard sheet: select a video node on the canvas, hit "+" → "Storyboard sheet", and the frames are sampled evenly across the clip and stitched into one large contact sheet, landing as a new image node you can feed into image-to-image, prompt reverse-engineering, or use as a reference
• 🧮 7 presets + custom grid: pick 4 / 6 / 9 / 12 / 16 / 20 / 24 frames in one click (portrait videos automatically switch to a portrait grid), or set rows and columns yourself (1–8 each, up to 48 frames)
• 🖼️ Up to 4K, no distortion: long edge up to 4096px, every cell keeps the source aspect ratio and is never upscaled; saving the project stores the sheet at full size
• 👀 Preview on open: the dialog immediately shows each cell's anchor frame with its timestamp and refreshes as you change the grid; when you generate, the sharpest, most distinct frame in each segment is chosen, avoiding black frames, blur, and near-duplicates
• ⚙️ Options: gutter 0 / 8 / 16, black or white background, numbered cells, and "also output single frames" (an extra pageable multi-image node you can extract frames from)
• 💳 0 credits: everything runs locally in your browser and the source video stays untouched

v1.39.0 · 2026-09-09 — New model: GPT Image 2.5 🖼️

• 🆕 GPT Image 2.5 launches with two entries: Flare (the default — everyday generation, lower latency) and Sunburst (the refinement tier — steadier multi-round edits, more precise local replacement), both selectable directly in the model dropdown
• 💳 Priced exactly like GPT Image 2: 1K / 2K / 4K cost 8 / 13 / 20 credits, Flare and Sunburst cost the same, text-to-image and image-to-image cost the same, and aspect ratio or reference image count never change the price
• 📐 13 aspect ratios + up to 16 reference images: 1:1 / 3:2 / 2:3 / 4:3 / 3:4 / 16:9 / 9:16 / 21:9 / 27:16 / 16:27 / 9:8 / 8:9 / auto (🔴 no 4:5 / 5:4 — those exist on GPT Image 2), with prompts up to 20,000 characters
• 🖼️ Gallery integration: GPT Image 2.5 gallery posts can be reused with one click to continue creating with the same prompt and settings

v1.27.0 · 2026-08-29 — New model: Gemini Omni Flash 1.1 🎬

• 🆕 Gemini Omni Flash 1.1 launches: Google's newest video model, supporting text-to-video and image-to-video, with prompts up to 20,000 characters
• 🖼️ Two image-to-video modes: Reference-image mode (up to 7 reference images) and First/Last-frame mode (first frame plus an optional last frame)
• 📐 4 durations, 2 aspect ratios, 3 resolutions: 4 / 6 / 8 / 10 seconds, 16:9 / 9:16, and 720p / 1080p / 4K (4K requires a PAYG / Starter plan or above)
• 💳 Fixed per-clip pricing, not per-second: 720p and 1080p cost the same — 79 / 105 / 132 / 158 credits for 4 / 6 / 8 / 10 seconds; 4K costs 184 / 210 / 237 / 263 credits; aspect ratio, reference image count and first/last-frame mode don't change the price

v1.25.0 · 2026-08-15 — New model: Grok Image 2 🖼️

• 🆕 Grok Image 2 launches: xAI's new image model (officially Imagine Image 2.0), built around tighter prompt adherence and cleaner in-image typography — strong on busy, text-bearing compositions such as posters, infographics and social graphics
• 📐 5 aspect ratios + long prompts: 2:3 / 3:2 / 1:1 / 16:9 / 9:16, with prompts up to 5,000 characters; it is text-to-image only for now (no reference images or image-to-image), with image editing to follow
• 💳 Simple pricing: 5 credits per image, the same for every aspect ratio, with no resolution or quality tiers; a render takes about 80 seconds in our testing
• 🔒 Security: generation logs hardened with redaction

v1.24.0 · 2026-08-12 — Canvas Skills drop 🎨

• 🆕 13 new Skills at once: 8 of them are fixed illustration recipes carrying 25 styles in total — Editorial Minimal / Fashion Colorblock / Soft Dream / White Couture / Dark Fashion / Oriental Poetry / Print Poster / Cinematic Narrative; no long prompt to write, just pick a style and name a subject
• 🖌️ 5 brand-new skills: Tiny Doodle IP (wobbly pen-doodle characters on lots of white space — attach a character reference image to keep the same figure across images), Paper Zine Poster (70–90% white space plus one saturated accent colour, freely combining 6 anchor positions × 7 accent colours × 5 paper tones), Photo Revival (upload a photo and get it redrawn as a hand-drawn illustration on white paper), Scene Paper Collage (your photo pasted onto handmade paper and rebuilt from simplified colour shapes), and Photo Note Panel (half the frame keeps the photo as-is, half is an abstract note panel with a short English title)
• 🗂️ New "Illustration" category in the Skills panel: illustration skills no longer crowd the "Creative" group; list cards now show a cover thumbnail with a large preview on hover
• ☁️ Sky Mirror upgrade: compositions grow from 4 to 7 (floating continent terraces, island-chain crossings, nine-layered sky worlds) and a new "scale" parameter (vast / epic / colossal / follow the theme) takes the same subject from restrained to monumental
• 🖼️ Clearer image-required skills: when you haven't attached enough images the button is disabled and shows how many are still missing (nothing is generated and no credits are spent), and going over the limit tells you only the first N are used; direct-output skills now share a single "Generate" button label
• 🐛 Fix: after dragging a file onto the canvas, the drop highlight could get stuck and only clear on a page refresh

v1.23.0 · 2026-08-08 — Seedance 2.5 🎬

• 🆕 Seedance 2.5 launches: generation length doubles to a freely selectable 5–30 seconds (previously 15 seconds on Seedance 2.0), at 480p or 720p
• 🌐 Full multimodal reference support: up to 9 reference images, up to 3 reference video clips (≤30s combined), and up to 3 reference audio clips (≤30s combined); prompts can use @Image1 to point directly at a reference image's role; a new adaptive aspect ratio joins the list as the default, and first/last-frame control is still supported
• 🔊 Generate-audio toggle: turn it off when you don't need a soundtrack or voiceover, for a silent video instead
• 💳 Two-track pricing: without a reference video, billing is by output seconds (35 credits/s at 480p, 79 credits/s at 720p); with a reference video the rate is lower but billed on input+output total seconds (22 credits/s at 480p, 48 credits/s at 720p), with a 5-second minimum

v1.17.0 · 2026-08-02 — MiniMax H3 (Hailuo 03) 🎬

• 🆕 MiniMax H3 launches: text-to-video and single-first-frame image-to-video, with fixed 2K output and freely selectable 4–15s duration
• 🔊 Native stereo audio: visuals, ambience, sound effects and speech are jointly generated, so prompts can specify where a sound comes from and when it occurs
• 💳 Billed by actual duration: 46 credits per second, or 184 credits for 4s and 690 credits for 15s; Reference-to-Video and video/audio reference inputs are not currently available

v1.16.0 · 2026-08-02 — Grok Imagine Video 1.5 🎬

• 🆕 Grok Imagine Video 1.5 launches: the original Grok video model remains available, while 1.5 is added as a separate choice with text-to-video and image-to-video, 480p / 720p output, and freely selectable 1–15s duration
• 🎥 Better motion and sound: 1.5 focuses on more coherent movement, more believable physics, and native ambience, sound effects and speech generated together with the video
• 💳 Billed by actual duration: 2 credits per second at 480p or 4 credits per second at 720p, with no 6-second minimum; a 2-second 480p clip costs just 4 credits

v1.13.0 · 2026-07-29 — Canvas video transitions 🎞️

• 🆕 Video transitions launch: box-select two videos on the Infinite Canvas and start "Video transition" from the action panel — the last frame of the first clip and the first frame of the second are taken automatically as the first/last frames, producing a transition clip that bridges the two naturally
• 🧭 Position sets A/B, easy to double-check: node position decides the order (top→bottom, left→right, same rule as multi-image references), nodes carry a "Video A / Video B" badge, and the action panel shows an "A last frame → B first frame" thumbnail pair you can swap in one click; black frames are skipped automatically when grabbing the bridging frames. The transition style description is optional (e.g. "flash-to-white", "camera-move bridge") — leave it blank for a smooth cross-over by default; a video model supporting first/last frames is picked automatically (the Seedance family) and can be switched by hand, with a selectable duration that defaults to the cheapest shortest tier
• 💳 Pricing: billed at the standard video generation rate of the model you pick — transitions carry no extra charge (e.g. Seedance 2.0 Fast · 480p · 4s = 80 credits)
• 🔧 Polish: transition results land back on the canvas node in real time and keep their connections; text in dialogs opened after a canvas box-select no longer shows a stuck selection highlight; the cost estimate now shows the real duration breakdown

v1.12.0 · 2026-07-29 — Video-to-script reverse engineering 🎬

• 🆕 Video reverse script launches: select a video node on the Infinite Canvas and start "Reverse script" from the action panel — frames are sampled at a set interval and analysed in a single batch by a multimodal vision model, producing a five-part storyboard script (overview / shot breakdown / lighting & mood / pacing / prompt), where the prompt section can be pasted straight into AI video models such as Seedance, Veo or Kling
• 🎚️ Four detail levels: Fast (1 frame every 2s) / Standard (1 frame per second) / Fine (2 frames per second) / Ultra (4 frames per second); in our testing a 10-second clip took about 48 seconds and came back with 10 shots on a per-shot timeline
• 💳 Transparent pricing: 6 credits base + 0.65 credits per frame, settled on the frames actually sampled (a 10-second clip on Standard = 11 frames = 14 credits); the estimate is shown before you confirm
• 🔧 Fix: script text nodes can now start a video generation directly — reverse, rewrite, generate, all without leaving the canvas

v1.11.0 · 2026-07-29 — Video upscaling (Topaz engine) 🔍

• 🆕 Video upscaling launches: powered by the Topaz professional video enhancement engine, with three levels — 1× (quality enhancement only, no resize), 2× and 4× — scaling width and height together (2× doubles both)
• 🖥️ Two entry points: hit "Upscale video" on a video card in your generation history, or select a video node on the Infinite Canvas and start it from the action panel; the result lands straight back in your history / canvas node
• 💳 Transparent pricing: 10 credits per second for 1× and 2×, 18 credits per second for 4×, billed on the input video's duration (a 5-second clip at 2× = 50 credits); the dialog shows a live estimate and settles on the actual duration
• 📦 Upload limits: up to 50MB per video, mp4 / mov supported

v1.10.0 · 2026-07-28 — Seedance 2.0 full power & Midjourney 8.2 🎬

• 🆕 Seedance 2.0 full upgrade (Fast / Pro / Mini): new reference video input (up to 3 clips, ≤15s combined) and reference audio (up to 3 clips, 2–15s each, must accompany a reference image or video); first/last frame can now be combined with reference material at the same time (previously either/or); generation length is freely selectable from 4–15 seconds (previously a few fixed tiers)
• 💳 Clearer pricing: with a reference video, billing is based on input + output seconds — the real duration is read automatically after upload, with the second-by-second breakdown shown before you generate
• 🔇 Generate-audio toggle: turn it off when you don't need a soundtrack or voiceover for a silent video instead
• 🆕 Midjourney 8.2 launches: the newest MJ flagship, 4 images per run, image prompts, style reference (--sref), and native --stylize/--chaos/--raw parameters.
• 🎨 Infinite Canvas: select a single video node to kick off video-to-video generation directly, with the same real-duration price breakdown

v1.7.0 · 2026-07-09 — New models & smarter picks ✨

• 🆕 Grok Imagine (xAI): text-to-image (6 variations per run, or a finer 4-shot Quality mode), image-to-image, and video generation (text/image to video, 480p/720p, 6–15s, up to 7 reference images)
• 🆕 Seedream 5.0 Pro (ByteDance): high-fidelity text-to-image and image-to-image with up to 5 reference images, 1K/2K output — great for posters and product shots
• 🖼️ Variant picker: when one run returns multiple images, flip through them right inside the result card and extract any variant as its own canvas node — downstream actions always use the one you're viewing
• ✅ Fairer content review: moderation is now far more permissive and precise — creative prompts (fashion, action scenes, celebrity-style posters …) no longer get wrongly blocked, while clearly harmful content stays out

v1.6.0 · 2026-07-08 — Annotate, sketch, create 🖊️

• ✏️ Annotation-driven editing: mark up an image on the canvas (pen, arrows, boxes, text) and run "Edit by annotation" — or feed the marked-up image straight into regular image/video generation
• ✨ Sketch to image: draw shapes freely on an empty canvas, select them, and turn your sketch directly into an image or video
• 🎯 Annotation export now matches exactly what you see: true colors, crisp arrows, no white borders
• 🎬 Multi-image to video: select several images and animate them in one run (first image becomes the opening frame)
• 🪟 Multi-window safety: opening the same canvas project in two windows no longer causes save conflicts — the second window opens read-only with a one-click "Take over editing"
• 💾 Canvas autosave is faster and far more reliable; favorites no longer disappear after a refresh

v1.5.0 · 2026-07-05 — A richer gallery 🖼️

• 📄 Every gallery work now has its own detail page with localized titles and descriptions — easier to share and discover
• 🔍 Better browsing: cleaner lightbox, improved mobile experience

v1.4.0 · 2026-07-04 — Infinite Canvas goes live 🎨

• 🧩 Flow Mode: an infinite canvas where you generate, connect and iterate — image generation, video, background removal, upscale, reverse-prompt, crop and more as connected nodes
• 🤖 Canvas Agent (beta): describe your goal and let the agent plan multi-step generations
• 📁 Canvas projects: save, reopen and manage multiple canvases (limits by plan)

v1.1.0 · 2026-06-26 — Safer payments 🔐

• 🛡️ Stronger payment security and account protection
• 🤖 Human verification at sign-up and email verification for a safer community
• 💳 Smoother checkout flow

v1.0.0 · 2026-06-22 — Launch 🚀

The first release of Genvizu.

• 🎨 AI inspiration gallery: browse, search, filter and favorite thousands of prompts and works
• 🖼️ Image / 🎬 video generation: multiple models (GPT Image, Nano Banana, Seedream, Flux, Seedance, Veo …), priced precisely by resolution / duration
• ✨ AI prompt assistant: translate, polish, improve, smart-variable detection, image reverse-prompt
• 💳 Subscriptions + credits: plan subscriptions, pay-per-use, automatic refund on failure