October 2026

Video effects: AI Muscle, AI Kiss, Tiger Hug, Puppy Hug, Face Punch, Drunk Dance, Explode, Giant Hand Pickup, Earth Zoom Into Room, Search Bar Reveal and Standing Split.

Meet Older Me, Couple Then and Now, Glow-Up Slider, Infinite Clones and Pet Becomes Your Twin turn your photos into short, shareable moments.

Photo effects: Pixar-Style 3D Character, Caricature Maker, GTA-Style Filter, Emoji Sticker Pack, Chibi Figure with Box, 80s Yearbook Photo, Pregnant AI Filter and AI Skinny Filter.

Make a photo talk or sing: Talking Photo reads your script, and AI Lip Sync follows your own voice clip or song.

Try it before you commit: AI Hairstyle Changer, Tattoo Try-On and Change Photo Angle.

Video Face Swap puts you into any clip, Multi Face Swap swaps every face in a group photo, and Video Colorizer brings black and white footage into full color.

For sellers and planners: Mockup Generator, Amazon Listing Images, Flat Lay Product Photo, Invitation Card Maker and Photo Collage Maker.

Holiday video effects: Christmas Card, Family Holiday Card (bring three to six separate photos together in one moving card), New Year Fireworks Duo, Thanksgiving Table Takeover, Rangoli Reveal for Diwali and 2026 Year in Photos.

Halloween is here: VHS Horror Cam, Group Costume Squad for two to four friends, Halloween Makeup, Halloween Cosplay and the Masked Stalker Selfie.

Holiday photo effects: Christmas Card, Holiday Polaroid Duo, Christmas Pet Poster, Diwali Portrait, Diwali Card and My 2027 Prediction.

Product Unboxed by You: a selfie plus a product photo makes your own unboxing ad, made for Black Friday.

12 new Sensei recipes for the season: a Halloween horror short starring you, a Diwali greeting pack, a holiday card and video card from your family photo, and a Christmas, Black Friday and New Year ad set from one product photo.

Recipes that sell: a product link to three UGC ads with a Black Friday countdown cut, a viral ad cloned for your product, a launch trailer, and a real-estate reel that dives from orbit to the front door.

Plus a week of influencer posts from one selfie, an old photo restored so you can hug your younger self, you as a boxed action figure, and a music video for your song. Every recipe shows the real result Sensei made with it at pixeldojo.ai/recipes.

Key pages now read natively in Spanish, Portuguese, German, Japanese, Italian, Polish, Norwegian and Traditional Chinese, from our comparisons to the genre pages and the free effects page.

A new PixelDojo Free page shows everything you can make on a free account, starting with 3 free video effects every day.

Pages no longer jump while you scroll on a phone. The notice strip at the top now slides away under the page instead of yanking everything up, and tool, effect and photo-effect pages stop shifting a moment after they load.

Headlines and text appear in their final font right away on slow connections, so the page doesn't reflow as you start reading.

What's New loads much faster: the latest updates are on the page, and older ones are grouped by month.

Links across the site that pointed to retired pages now go straight to the right page, and the menu sends you to a page you can actually open when you're signed out.

Pick and preview your photo on any Video Effect or Photo Effect before you have an account. Sign-up comes when you press Generate, and you land back on the effect with your photo ready.

The home page, the pricing page and every effect now say it up front: a free account gets 3 video or photo effects a day, no card needed.

Pricing shows how much the prepaid terms save, and the refund policy sits right under the plans.

Pages no longer start dimmed while they load, you see one notice at a time instead of a stack, and comparison pages use the main menu.

Recast has a clearer name: Character Swap. Same tool, same price, and old links take you straight there.

Recipes are easier to find: they are in the menu, under Studios quick starts, and on your dashboard.

Every recipe page has share options: copy the link, your phone's share sheet, X, Facebook, WhatsApp and Reddit. Links show the example in the preview.

When a Sensei reply comes out right, press Publish as a recipe under it. Sensei saves it if you have not yet, then the publish form opens. Every page is still reviewed before it goes live.

Recipe pages and cards show who made them, how many people ran them and how many remixes they have.

Ask Sensei for any Video Effect, Photo Effect or AI Tool by name: "AI Hug these two photos", "make me an action figure", "ghibli this", "remove the watermark". It finds the effect, runs it on your photos and shows the result in the chat.

Our MCP server and API can run them too: pixeldojo_effects lists every effect with its price, and pixeldojo_effect runs one by name on your photos, so Claude, ChatGPT or Cursor can make an AI Hug clip for you.

Same price as on the effect pages, and only charged when it works.

FLUX 3 Image Klein is here: the fast FLUX 3 model for images and edits. Type a short idea and it comes back as a fully composed scene, with any words in quotes set the way you wrote them.

Edit and combine up to 10 reference images: keep a product and move it to a new scene, or put people and objects from several photos together. 0.5 credits an image at 1K, 1 credit at 2K.

Klein is part of Relax, so every plan can run it unlimited with no credits. It also works in Canvas, the API and our MCP server.

New prompting guide with real examples and copy-ready prompts: pixeldojo.ai/guides/flux-3-image-klein-prompting-guide

Film Studio has a Final clip quality: clips render in 1080p and the finished film exports at 1920x1080, or 1080x1920 for vertical. Pick it in the film's settings; the price per clip shows before you render.

Faceless Studio has a Final 1080p scene option. Use it on any scene and the whole episode exports in Full HD, captions included.

The trailer, micro-drama and explainer pages now show films made start to finish in the studios at 1080p, exactly as they exported.

Our affiliate program now pays cash every month: 25% of every subscription payment from customers you refer, for 12 months each. Join in one click on your Referrals page and get your own link.

Your Referrals page shows your clicks, customers, pending and ready commissions, and payouts. Payouts go out monthly once you have $25 ready, by PayPal, Whop, or Pixel Dojo credits.

Creators get paid $1.50 per 1,000 views on videos they make with Pixel Dojo and post on TikTok, Instagram, YouTube or X, from 5,000 views, up to $300 per post. Submit the post on your Earn page.

Everything lives on one Earn page, linked from the menu: your affiliate link and earnings, your creator posts, and your credit referral link. Each program has its own page explaining how it works.

Credit referrals still stack on top: you and a friend each get 100 credits when they subscribe.

Every plan now includes unlimited Relax runs on Qwen Image 2.1 and MiniMax H3 Turbo. Pick Relax next to Generate and the run uses no credits.

Relax works in Video Effects and Photo Effects too, so you can try effect after effect without spending credits.

When the Relax queue is busy, we email you when your result is ready, so you can close the tab.

Publish a saved Sensei recipe as a public page with an example of what it makes. Anyone can open it and run it with their own product or idea, and the page credits you.

Every page is reviewed by a person before it goes public, and you can unpublish yours any time from the recipe picker.

Character Swap puts you or your character into a short video in one step. Add a 2 to 30 second clip you have the rights to, pick one of your Character Studio characters or upload a photo, and press Swap. The scene, camera, moves and sound stay the same, and the lead becomes your star. 3.4 credits per second, or 4.9 on Professional.

Swapping a character into a video now asks once for your agreement that everyone in it is an adult who agreed to be in it, the same as Face Swap.

A video you upload or paste a link to now shows its preview in the composer on every video tool, instead of a blank tile.

One-click starts for a movie trailer and a vertical micro-drama in Film Studio. Each opens the studio with the format, length, look and a sample idea filled in, and trailers come out cut like a trailer, with quick shots instead of long ones.

A UGC ad start in Marketing Studio, an explainer start in Faceless Studio, and two new Sensei recipes: a product ad from a store link, and an AI influencer post (a four-photo carousel of the same creator, with a caption).

New pages for each of these with a real example we made, what it cost and how long it took.

The menu bar now shows what you came to make: Sensei, Studios, Effects, Tools, Models and your Library, the same on every page. Hover any of them to see what is inside.

Studios lists all six studios plus quick starts for Character Swap, music videos and Canvas. Effects holds Video Effects, Photo Effects, Presets, the prompt gallery and the community gallery.

Every model now sits under Models, which opens on the ones you use most. Search with ⌘K and your pinned tools work as before.

The Library matches the rest of the site: a cleaner header, tabs that show how much is in each section, and larger tiles for your characters, elements and LoRAs with the name on the picture.

Search the Library overview to filter your characters, elements, LoRAs, prompts and recent media by name at once, with a one-click hand-off to the full media search.

A recent image or video on the overview now opens straight in the viewer, and the tool that made it shows on hover.

The Media and LoRAs tabs lost their duplicate titles and heavy shadows, and an empty Library now shows where to start instead of five empty boxes.

The home page now starts with what you want to make: one-click video effects, photo effects and one-job tools like headshots, photo restoration and background swaps, each a tap away.

The six studios and Sensei come right after, followed by reviews and plans, so it is quicker to see what Pixel Dojo does and what it costs.

Video Analyzer now uses a stronger vision model. Shot descriptions are more accurate, and it says when text is too blurry to read instead of guessing.

Long single takes are no longer described from their first frame alone. The breakdown now follows the shot through its middle and end, so a pan or a reveal makes it into the description.

Pricing follows the video: 3 credits per started minute covers up to 10 shots a minute, and fast-cut videos add 1 credit per 5 extra shots. The maximum is held when you start and the unused credits come back when the analysis finishes, so most short clips still cost 3 credits.

WAN Character Swap videos from the API and AI assistants now always arrive as a PixelDojo link, and a finished swap no longer shows as "still processing" forever.

Photo Effects: 30 one-click looks for your photos, like Ghibli Style, Action Figure, Sticker, Aging and Baby filters, Magazine Cover and Plush Pillow. Upload a photo and press Generate. Free accounts can use them as part of their 3 free effects a day.

AI Tools: one-job tools for the edits people ask for most. Remove watermarks, objects, text or crowds, restore and colorize old photos, unblur, uncrop, relight, make a headshot, put products on white or in a scene, translate the text in an image, and wipe subtitles off a clip.

OmniHuman is now OmniHuman 2.0: more natural lip sync, expressions and gestures from one photo and a voice clip, and clips up to 2 minutes long (up from 35 seconds). Same price, 5 credits per second of audio.

Use the prompt to direct the scene, the movement and the camera, in English, Chinese, Japanese, Korean, Spanish or Indonesian.

The new Alternatives section compares Pixel Dojo with the image, video, avatar and voice tools you might be weighing. Every comparison runs a real brief on our models, shows the results with their prompts and credit costs, and links its sources.

Video Effects is here: 283 ready-made effects like AI Hug, AI Flying, 360 Rotation, Earth Zoom In and Old Photo Animation, plus product ads, real estate walkthroughs, anime looks, Halloween and holiday effects. Upload a photo, press Generate, and get a 5 second HD video. No prompt to write.

Two-photo effects bring two people into one scene, and Smooth Transition morphs one photo into another. Find them all under Video Effects in the menu.

Try them free: every account gets 3 Video Effects a day, with a small PixelDojo watermark. Subscribers run as many as they like with their credits, with no watermark.

P-Video 2 Pro now takes references. Add up to 9 images, 3 videos and 3 audio clips to carry a face, a product, a motion or a voice into your clip, and keep your start and end frames at the same time. From 1.25 credits per second.

Add one song or voice clip and tick "Use as soundtrack" to get a clip that lip-syncs to your exact audio.

Film Studio music videos now start every shot on your approved storyboard frame while singing your song, and every take plays your exact audio.

Swap a reference image in place with the new replace button on its tile. It keeps its number, so @Image1 in your prompt now points at the new picture.

Drag reference images to put them in a different order, or use the up and down arrows in the full list when you have lots of them.

Your prompt keeps up with you. When you reorder or remove a reference image, tags like @Image3 or "image 3" update so each one still points at the same picture.

The full Qwen Image 2.1 model now costs 1 credit per image at 1K (was 0.5). 2K stays at 1.5 credits, and Qwen Image 2.1 Turbo is unchanged at 0.25 credits at 1K and 0.5 at 2K.

Films with see-through, glowing or ghostly characters no longer get a clothing warning on every clip. The clip check now treats those looks as the effect you asked for, not a costume mistake.

The review of your finished film now reads what each shot is meant to show, so it stops flagging a change of clothes your script asks for and points only at real slips, like a jacket that comes back on with no reason.

Add a classic lower third for interviews and talking heads: a bar in your brand color beside a bold name and a lighter title. On widescreen video it sits in the traditional spot at the bottom.

Add a countdown that really ticks down on screen, like "Ends in 23:59:59", right under your offer badge.

Add a price tag sticker with your new price and the old one struck through, like $49 instead of $79.

LoRA training with a larger set of photos could stop with "signal timed out" before it started. Fixed on every trainer, so the same zip goes through now.

Brand Overlays now adds a hook title over the first seconds, a callout that slides in for a benefit or a name, a review card with stars, and a progress bar that fills as the video plays. Mix them with the offer badge and end card in one go.

Every overlay stays clear of the TikTok, Reels and Shorts buttons, and the offer badge waits for your hook to finish so they never overlap.

Brand Overlays adds an offer badge like "50% OFF" that pops in on cue, and an end card with your logo, headline, button and web address, in your brand color. The words come out exactly as you type them.

Ask Sensei to "add an end card with my logo" or "put a 20% off badge on this", or call pixeldojo_overlay from your own AI agent.

Video Autocaption now lights up each word as it is spoken, in bigger type placed above the TikTok, Reels and Shorts buttons, and it is faster.

Pick a caption style (Pop, Highlighter or Clean) and any highlight color, in Video Autocaption and in Faceless Studio episodes.

Ask Sensei to caption a finished video ("add captions to this in my brand color"), or call the new pixeldojo_caption tool from your own AI agent.

The clip check now looks twice before saying a character's face changed or does not match their photo, so you see fewer warnings about clips that were fine.

A shot that continues the previous clip now gets a face check too. If someone in it no longer looks like their photo, the storyboard tells you before you render the clip, so you can paint a fresh frame instead.

Editing images on Qwen Image works again, now on the newer Qwen Image 2.0 engine at the same price. Plus and Max edits had been failing since October 10.

Qwen Image 2.1 is here in two versions: Turbo for fast native 2K images, and the full model for the most detail and the sharpest lettering.

Add up to 10 reference images to edit a photo or combine people, products and scenes into one shot.

Train your own Qwen Image 2.1 LoRA from a zip of photos, then use it in Qwen Image 2.1 with one click from My Trainings.

The Qwen Image 2.1 prompting guide has real examples you can copy, from posters with exact type to product shots and your own LoRAs.

Film Studio now writes each character's accessories and, when there is no reference photo, their hair into their look, so glasses, hats and jewellery stop coming and going between shots of the same scene.

When the last clip of a film finishes rendering, Film Studio saves every clip in it for you, so a finished film never loses clips to the 24 hour draft limit.

If you leave unsaved clips or frames behind, Film Studio emails you a few hours before they are deleted, with a link straight back to the film.

Export stays the main button while the editor's cut is waiting. Applying the cut is optional, and the line under Export says when it is not applied.

Before a batch of frames or clips, Film Studio checks your current credits. If they will not cover the whole batch, it tells you and renders only the shots they cover, instead of stopping partway.

The review of a finished film no longer reports black frames that are not there.

Seedance now accepts high frame rate phone videos, like 120 fps iPhone clips. We convert them to 30 fps for you automatically instead of asking you to re-export.

When Seedance can't fetch one of your references in time, it now tries again on its own instead of failing the run.

When you run out of credits in Video Studio, the message now links straight to buying more.

When a model turns a Character Studio shot down, the card now offers a one-click retry on Seedream 5 Pro with the same prompt and attachments. The blocked shot is refunded, and the button shows the price.

A shot that fails in Character Studio now says why, instead of showing "Generating" forever.

Character Studio's model list drops Seedream 5 Flash, which turned down most character shots. Seedream 5 Pro and the other models are unchanged.

Refine after Enhance now always turns your instruction into a new version of the prompt, instead of sometimes clearing what you typed and changing nothing.

If the assistant needs to ask you something before refining, you now see its question, and your instruction stays in the box.

Affiliates now earn 50% of every new subscriber's first month, paid once, on Lite, Pro and Pro Max.

The Affiliate page has a promo kit: ready-made links to every studio and top model, videos you can repost, and copy you can paste.

Subscriptions started from an affiliate's link are now credited to that affiliate.

Video Studio now marks the models that often decline a prompt like yours and offers a one-click switch to one that usually makes it, before you spend anything.

Kling models now tell you right away when a request is outside their content rules, instead of failing after a long wait. Nothing is sent and you aren't charged.

When Seedance 2.5 treats a prompt with a reference video as an edit of that video, it now tells you how to fix it (set the length to Auto, or use Video Edit) instead of showing a technical error.

Character Studio now warns you before you generate when the model you picked will likely turn the shot down, and switches you to Seedream 5 Pro in one click.

Expand references now uses Seedream 5 Pro when the usual model would refuse your photo, so you get four new angles instead of an error. The button always shows the price.

Characters with a single photo get a quick prompt to add more or make four angles, since characters with three or more photos keep their face far more reliably.

A new character's studio now offers four one-click starter scenes, so your first shot is a click away.

When Sensei's review finds a problem with a take, it now shows you the take, its score and what's off, and offers a retake with the price, instead of redoing it on its own. You only pay for retakes you want.

Plan quotes now cover one take of each shot, with no retakes built in, and Sensei tells you up front when your balance is short of a plan so you can trim it or top up first.

Sensei no longer keeps spending on a new project you haven't approved yet, and a film's charges now count toward your approved credits as each clip renders.

Ask Sensei for just a prompt and you get the prompt, with nothing rendered.

WAN 3.0 shots that use both a start frame and your reference photos now render instead of failing.

A step that stops answering no longer stalls a whole reply, and a refused first message no longer leaves an empty chat in Recent.

Film Studio's frame check stops redoing a frame once it fails twice for the same reason, and no longer pays for a redo just because the shot size or expression is off. It flags the frame for you instead, so you decide.

Export now saves any unsaved clips for you before it builds the film, so a finished storyboard is one click from a finished film.

Frame and clip checks now watch for clothing that changes between shots, and every clip is asked to keep each person and their outfit the same from start to finish.

When the script writer can't write an idea, Film Studio now tells you why and leaves your film as it was, instead of saving the refusal as your script.

When a script can't be broken into shots, the message now explains the likely reason and what to change.

Film Studio clips no longer bring in characters who aren't in the shot: a character waiting elsewhere in the scene used to walk into other shots, sometimes as a copy of the lead.

WAN 3.0 clips in Film Studio now start exactly on your approved storyboard frame, so characters, faces and composition stay more consistent from shot to shot.

WAN 3.0 clips in Film Studio now stay in one continuous shot instead of cutting to a new angle partway through.

Film Studio clips on every engine now stick to your approved storyboard frame, instead of rearranging the room or cutting away to show parts of the set the frame left out.

Setting a shot longer than 10 seconds now shows a tip: long clips fail the clip check about twice as often, so two shorter shots usually look better.

Seedance 2.5 and Marketing Studio no longer stay on 480p draft after you leave the page or reuse a draft's prompt. It comes back at 720p, so you only get a draft when you pick one.

Seedance 2.5 drafts are now made only when you pick one yourself on the Seedance 2.5 page or in Marketing Studio. Sensei, connected AI assistants and the API always render full-quality videos.

Nano Banana 2.1, Google's newest image model, is now the default in the Nano Banana generator. Sharper layouts and text, more consistent characters and products, and more precise edits, at 1K, 2K or 4K.

Nano Banana 2.1 costs less than Nano Banana 2: 1 credit at 1K, 1.5 at 2K and 3 at 4K. Nano Banana 2 is still in the model picker if you prefer it.

Character Studio, Film Studio frames, Character Sheets, Consistent Characters, Image Relighting and Reality Polisher now use Nano Banana 2.1 wherever they used Nano Banana 2.

Nano Banana 2.1 is also on Canvas for generating and editing.

A new Nano Banana 2.1 prompting guide shows how to brief posters, infographics, consistent characters and precise edits, with real examples and prompts you can copy.

The credits you approve on a Sensei plan are now the total for the whole project, not a fresh allowance for every message.

The plan shows how much of your approved credits it has used, and Sensei stops and asks before going past it.

Need more? One button adds just enough to finish, and you see the amount before you press it.

The Flux 2 LoRA Trainer can now train on the full-size Flux 2 Dev, for the most detail and likeness, at 110 credits per 1000 steps. Flux 2 Klein 9B stays as the faster option.

The trainer asks up front which model you're training for and shows where your LoRA will run, how long it takes and what it costs.

Pick a LoRA on the Flux 2 page and it switches to the model that LoRA was trained for. Your library now labels each one Flux 2 Klein 9B or Flux 2 Dev.

Refunds for failed generations are now paid exactly once, even if something goes wrong at the worst moment.

When a model is briefly unavailable, the message now tells you exactly what happens to your credits, refunded now or automatically if the job fails.

Sending the exact same request twice in a row no longer charges you twice.

API: send an Idempotency-Key header and a retried request returns the original job instead of charging again.

API: a quick result that hit a storage hiccup is now delivered on your next status check instead of being refunded hours later.

API: every model in the catalog shows the price you are actually charged.

Stop now stops Sensei right away, even mid-sentence, and a reply that stalls is cut off and retried instead of spinning forever.

Notes you send while Sensei is working are read once, never dropped or repeated.

Films, videos and episodes Sensei makes now count toward its spending limit, so a chat stays within the budget you approved.

If we update the site while Sensei is replying, it picks up where it left off without repeating itself.

A finished image that fails to load on a shaky connection now retries and offers a Retry button, instead of disappearing from your history.

A brief hiccup on our side while a video renders no longer marks it failed. We keep checking, and you see it as soon as it's done.

Double-clicking Generate no longer sends the same job twice.

Pages and your dashboard load faster, and background tabs check on jobs less often, then catch up the moment you come back.

No more false "new version available" banner on some pages.

Your results grid stays smooth while a video renders, instead of redrawing every few seconds.

Flux 2 Dev + LoRA: bring your own Flux 2 Dev LoRAs to the Flux 2 page, stack up to three, and add up to four reference images. 1 credit at 1 MP.

The first time you use a new LoRA it takes a few minutes to get ready. If you try before it's ready, you're told to come back and your credits are returned.

On your phone, announcement and sale banners tuck away as you scroll, and the dashboard is about a quarter shorter, with guides you can swipe through.

Jump back in shows your latest result from each of your favorite tools instead of a plain icon.

Community Spotlight now shows a wider mix of what people make: scenes, products, signs and video, not just portraits.

On Sale cards fill the row and say how many tools are discounted, and Canvas joins the Studios row.

The editor's notes now say what changed in plain words, without their own runtimes that could disagree with the cut's length.

While the editor's cut is waiting, Apply edit is the one highlighted button, and Export steps back until you've decided.

Shot cards keep the Edit button in the same spot, with the clip's length and trims on their own line.

A character a shot only mentions off screen (someone another character is looking at) no longer gets drawn into that shot.

The storyboard's price now includes the frames the frame check may redo, so you see the full range before you render.

After casting a character, the shot editor shows their outfit once instead of twice.

Casting a role no longer offers a character who is already playing someone else, and Regenerate shows its price.

The frame check now judges a frame as the opening moment of its clip, so it stops redoing, and charging for, frames that were already right.

New storyboards describe a look toward someone off screen as a direction, so no stranger turns up at the edge of the frame.

A clip that just started rendering no longer shows the Stuck button.

When the editor's cut is ready, the step guide points you to it.

Every film now shows which step it's on (Script, Frames, Clips, Export) and the one thing to do next.

Shot cards are simpler: they say how long an unsaved frame or clip has left, and when the frame check spots a problem, the card says exactly what it found.

Settings shows your current frame and clip model with a Change button instead of a long list, and Time period replaces the old Era chip.

Coming back to Film Studio puts your films first.

Using your own image or video for a shot is one Use my own button, and the shot editor separates what the next render uses from free edits to the cut.

Starting a new film is clearer: the script page shows the steps with Break into scenes at the top, the storyboard tells you who to cast first, and making a character's portrait shows its price.

Film Studio now has an editor. Once your clips are in, it watches the whole film and makes the cut: tighter starts and ends, a little extra speed on a stunt that reads slow, and a shot that doesn't work left out when the next one tells the same moment. You see the before and after length and the editor's notes, then press Apply edit. Nothing changes until you do, and Undo puts it back.

The frozen frames a clip often opens on are found and cut, so shots no longer hiccup at every cut.

Spoken lines are never trimmed: the edit keeps every word and the action around it.

Films made by Sensei get the same edit before the final cut, and the finished reply says how long the film really runs.

Clip trims can now go longer than two seconds, and each shot can play faster or be left out of the film from its Clip tab, so a shot that lands early no longer holds on screen.

Assemble & export is now three clear steps: Music, Look and Export. You can see at a glance whether your film has music, compose a score or upload your own in one place, and read exactly what the export will include.

Export waits until every clip is saved instead of failing, with a one-click Save all clips right in the warning, and an older export is marked when the film has changed since.

Sensei and all of our studios now sit at the top of the main menu, so the fastest way to make something is one tap away.

On your phone, the menu opens on studios and categories that fit on one screen, with your recent tools right below.

Each category lists the newest models first. Earlier versions you still love are right below under Older models, nothing removed.

Model names are consistent everywhere: FLUX.2, Qwen Image 2, Veo 3.1, P-Video, Z-Image Turbo, WAN 2.7 Image and more.

Tutorials, What's New and the Fast FLUX Trainer are now in the main menu, and My Library is always there, even before your first generation.

The menu works fully from the keyboard and is easier to tap on touch screens.

Your film's music now rises between lines and dips smoothly under dialogue, instead of sitting at one low level the whole way through.

Clips no longer add music of their own, so the film's score is the only music you hear.

Your videos now fill their card edge to edge. Hover to preview a clip, then click to watch it full screen with sound.

On your phone, results show one per row with nothing underneath: just a Save button and a menu on the picture, with Edit, Upscale, Download and the rest one tap away in the menu.

The prompt and buttons on your results are easy to read in light mode again.

A generation that fails now offers Try again or Reuse prompt right on the card.

You can move through your results with the keyboard, and Enter opens one full screen.

Images in your history no longer get stuck loading after you refresh the page.

The Email preferences link at the bottom of our emails, and the Unsubscribe button your inbox shows next to our name, now turn those emails off instead of showing an error.

The Email preferences page lists every kind of email we send: getting started tips, news and sales, and offers to come back. Turn any of them on or off, or unsubscribe from all of them in one click.

When Sensei plans a film, the film now lives right in the chat: a storyboard that fills in as frames render, live progress, a Start production button, and the finished film playing inline. No more jumping to another screen.

Sensei now casts your saved characters as they are instead of making copies of them, and the shot planner sticks to the cast you chose, so each character appears once, in one outfit. Saving the same photos or props again no longer adds duplicates to your library.

Film Studio spots props that show up in several shots (a weapon, a jar, a car) and offers to lock each one in with a reference image before your frames render, so it looks the same every time. Sensei does this for you automatically.

Characters keep the gender from your description or their photo instead of a guess from their name.

When the clip check flags a clip, Sensei renders it once more before cutting your film, and tells you about any that still need a look.

If the site restarts while Sensei is producing your film, it now picks the same film back up instead of planning and paying for a new one.

A retro stripe lights up while Sensei thinks and while your images and videos are generating across the site, so you can always see that something is happening.

Sensei has a smarter brain for planning your project, picking the right tools and reviewing every take, and it asks for your product photo when an ad needs the real thing.

Sensei's thinking fee now only covers what it costs us to run, so most replies cost a fraction of a credit. Images and videos cost the same as always.

Casting a saved character with @ no longer also pulls in a different saved character who shares their first name, so your shots stop showing the same person twice.

Replacing someone in a film's cast works again instead of failing with an error.

While Film Studio writes your script or plans your shots, the progress now sits right on the prompt box with a moving retro stripe, and the stripe fills up as shots land.

Film Studio has a smarter writer and director. Stories make more sense on one watch, and storyboard frames and finished clips get a sharper check for wrong faces, stray people and broken hands, with far fewer false alarms. The check on your finished film also catches a car that changes model, people who swap seats, and outfits or props that change mid-scene.

Every character in a film we write for you is now clearly an adult, so the safety check no longer stops a film partway through.

A character in your film no longer borrows the face of a saved character who happens to share their name. Your saved characters still join when you name them in your idea or add them to the cast.

When a clip's generated soundtrack gets blocked, Film Studio now renders that shot again without sound instead of stopping your film.

A clip's "check clip" warning now stays after you save the clip, and a clip you save while its check is still running gets the result too.

Films no longer cut every couple of seconds. Each shot now plays its whole action, so a door opens and someone walks through it, and a line gets its reaction.

Scripts tell the story through what the characters say and do, never through signs, notes or narration the video can't show, so the twist actually reaches the screen.

Films are planned around the story's beats, about one shot every six seconds, with quick back-and-forth cutting saved for conversations. Horror holds longer on the dread, comedy holds on the reaction.

A character the story keeps hidden no longer shows up in the background before their big reveal.

Making a trailer or promo through the API or an AI assistant? Ask for trailer pacing and most shots stay on screen for a quick one to three seconds.

Films you already started are fixed too: the next export plays every shot in full.

Hand Sensei 3 to 5 photos of someone and ask it to save them as a character. It goes straight into Character Studio, ready to use as @Name anywhere. Products, outfits, locations, props and styles work the same way.

AI assistants connected to PixelDojo can now list your saved characters, add new ones from photos, and save a result to My Media when you say "save this". Everything passes the same safety checks as saving on the site.

The @ menu shows a thumbnail next to every saved item, and Film Studio and Video Studio now use the same menu. In Video Studio, picking someone from @ adds them to your references for you.

Sensei checks your film's real progress before telling you how far along it is, instead of guessing.

Type @ in any prompt box to pick a character, product, outfit, location, prop or style you saved in Character Studio. We send their photos to the model for you, so @Mara on a beach looks like Mara. Works on every image and video tool, in Canvas and in Sensei.

If the model you picked can't use reference photos, we tell you before anything is charged and suggest one that can. Saved places and props with a description still work there as text.

Once you approve a film or music video, Sensei keeps production moving until the finished video is ready, even if you close the tab. No more saying "continue" to nudge it along.

Ask Sensei to redo a shot of a finished film or music video: it renders a new take of that shot and cuts a fresh video with it.

A background tab only says Sensei needs your OK when it really does, and the Working timer under the chat keeps time while you're in another tab.

Every image and video Sensei makes now plays from PixelDojo's own storage, so Save works on every take and links no longer expire in the middle of a chat.

Vertical stays vertical: when Sensei animates a 9:16 still together with your product photos, the clip keeps the still's shape instead of coming out widescreen.

If Sensei's final cut can't be stored with us, you aren't charged for it and Sensei can simply cut it again.

FLUX 3 Image: photoreal images up to native 4K, edits that combine up to 10 reference images, and web grounding that looks up real products and places before it draws. From 1 credit per image.

Ideogram 4.5: posters, logos and packaging with text that reads right, plus plain-language edits. Turn on Precise Edit to change one thing and keep everything else exactly as it was. From 1 credit per image, and quick Draft edits at half a credit.

Both models are in Canvas, on the API and MCP, and each has its own prompting guide.

The home page shows all six studios, Prompt Studio included, and names the models behind every image and video lane, each linked to its page.

Drag the before and after slider on the home page to see what upscaling does to a real photo.

Each Sensei starter recipe now shows what it made when we ran it: a 15-second ad, a UGC testimonial, a product photoshoot, a short film, a music video and a poster with exact text.

FLUX 3 Image and Ideogram 4.5 lead the model tabs on the AI Image page and appear in the pricing table.

The LoRA page shows one trained LoRA in six different looks, plus a ready-made AI influencer set.

Copy any example prompt on the Skills page for ChatGPT, Cursor or Codex, try the interactive URL-to-ad demo on the AI Ads page, and read the full per-model price table on your phone.

Sensei now makes posters, logos, labels and packaging with exact text on Ideogram 4.5, and uses FLUX 3 Image when you ask for native 4K.

Real examples everywhere: a six-view character sheet on the Character Studio page, three full music videos to play on the AI Music Video page, and a vertical faceless episode, Why Do We Yawn?, on the Faceless Video page.

Opening a video on a page no longer jumps you back to the top.

See every tool now lists all of them: the All AI Tools page shows every generator, editor, upscaler, trainer and studio, grouped like the app's tool menu, with a link to each.

The home page, AI Image and AI Video pages now show the range: comics, manga, claymation, watercolor, posters, product shots, anime and more, each labeled with the model that made it. Click any example to open that model.

Same cast, any style: our Prime Time stars redrawn as anime, claymation, watercolor, film noir, pixel art, 3D animation, an oil painting and a pop-art comic, all from one set of Character Studio references.

The AI Music Video page shows all nine looks, from Rap video to K-pop, on the same band.

The AI Image and AI Video pages now have a tab for every model, 21 image and 20 video, each with examples in different styles and a link to that model's page.

Ask Sensei for a character by name and it finds them in your Character Studio, or among the photos you've sent it before, and uses them as the reference.

Starter recipes like "15-second product ad" now open on the first try.

Typing "approved" or "go ahead" under a plan approves it, the same as pressing Approve.

Sensei replies in the language you write in, knows where LoRA training lives, picks the right model for what you ask without you naming it, and only says it's working when a job has really started.

Sensei's clip reviews no longer time out, and the final cut is stored with us so it plays and downloads reliably.

In Video Studio, type @ and the first letters of an attached character or product to pick it from a list. Tapping a tile on the rail inserts its tag, and your cards show the prompt as you wrote it.

Sending a voice track from Seed Audio to OmniHuman now loads its length and price every time.

Finished Marketing Studio videos play from our own storage, so links keep working.

AI assistants checking a campaign now see where each image and video stands, not just a count.

Sensei has its own page: see how your AI creative director plans and prices every shot, checks every take, takes notes mid-run and saves recipes. Powered by Grok 4.7, part of every plan.

The home page now opens with a new film: our cast switching art styles on every cut, then a tour of the models, studios and Sensei. The Sensei promo plays wherever Sensei appears.

Marketing Studio's page shows how your own AI agent can run a whole product launch with Pixel Dojo's tools over MCP.

Film soundtracks now stay with your film, so you can re-export any time without making a new score. If an older film's soundtrack has expired, the export tells you how to fix it.

A four-part video series shows Sensei at work: give it a product link, approve the priced plan, watch it check its own takes, then save the run as a recipe for your next product. It plays straight through, opens on your first visit to Sensei, and is one click away from the Learn Sensei button and the Tutorials section.

Redo frame and Redo clip in Film Studio work again after you edit a shot or recast a character. For about a day they stopped with an error, without charging you or changing the shot.

Recasting a character as a new version of themselves, such as Brian as Brian (80s), no longer writes the new name twice in shot descriptions.

Sensei now checks that every image it's about to use really exists before it starts a job. If a link is wrong, it finds the right file in your chat or Library, or asks you for it, and nothing is charged.

Videos that start from a still and also use your reference photos now render instead of stopping with a settings error.

When a step needs the image upload agreement, it opens right away, and Sensei carries on as soon as you accept.

Sensei chats open far faster: results load as sharp previews instead of full-size files, a single result now shows big, and videos start when you scroll to them.

A plan waiting for your OK folds out of the way as you scroll the chat, with Approve still one click away on its header.

Recraft V4.1, Hunyuan Image 3, ImagineArt, PonyXL, WAN 2.7 Image, WAN 2.6 Image, WAN 2.2 Image, WAN 2.6, WAN 2.2, WAN 2.1, P Video and P Video Avatar each have their own page now, with examples we made and the exact prompt behind every one. Tap Use this prompt to start from any of them.

Model names on our pages only link where there is a page to see, so a click never lands on a sign-in screen.

WAN 2.6 Image text-to-image now reliably finishes with your image. About half of generations were failing and refunding, and they now render in a few seconds.

A result that does not match your request, such as the wrong size, is now refunded instead of shown to you.

Film Studio now casts every named character before it draws any frame, so each person and pet looks the same in every shot. Cast from your own photos, a short description, or a character you made in Character Studio.

AI assistants connected to PixelDojo can cast your film too: they list who still needs a face and cast each one from photos, a description, or your Character Studio library before production starts. They can also swap or remove a cast member until that person's shots are drawn.

Picking a different character for a part now replaces the old one in every shot, with their outfit and voice carried over, instead of putting both faces in the scene. Each cast member on the storyboard also has Replace and Remove buttons.

Casting one character no longer touches a similar name: casting the Woman leaves the Old Woman alone, so each person keeps their own face and still gets asked for one.

Frames and clips now land on your storyboard even after you close the tab. Anything that finished but expired before reaching you is refunded automatically.

The frame checker now compares every storyboard frame with your cast's photos and catches the wrong person, the same person twice, a missing or extra person, and broken hands. Each redo fixes the specific problem it found, with up to four tries per frame.

Films now cut like trailers: action shots show only their best first seconds, faces are framed to stay recognisable, and every clip and the finished film get checked for drifting faces, morphing hands and bad cuts, with notes on what to redo.

Film Studio: type @ in your idea to pull in a character from Character Studio, or click a cast member to drop their name into the prompt.

Film Studio frames now render on Seedream 5 Pro for sharper, more lifelike faces, and the frame check now rejects takes where the mood or shot size is wrong, like a smile in the middle of a chase.

Film Studio keeps the cast you attach: the script uses their names instead of inventing new characters, and a name like "Pamela (80s)" now matches PAMELA in the script, so there is nothing to recast on the storyboard.

Film Studio: Generate all frames now finishes the whole film, waiting its turn instead of stopping after the first few. Close-ups also come out as real face close-ups, with only the people the shot is about.