Changelog

A major release across Plumar TV, The Vision studio, and the partner program — with a 24-hour welcome gift for every new director.

Plumar TV opens with a full onboarding flow and a 24-hour gift so you can step onto the set the moment you arrive. Shot Director now follows a true cinematic standard, and Seedance 2.5 runs at full power — with Kling 3.0 stretching video generations up to 15 seconds instead of 5.

Inside The Vision, the first-visit walkthrough is smoother, nodes hold their place when you collapse a group, cast cards live in their own character library, and the depth map is now ours. Every generation shows up live in the notch — no more averaged numbers, no lost progress.

The partner program gets named links like plumar.ai/valeria, period-by-period earnings, and a clean partner card. Payments are unified in rubles through CloudPayments, gifted frames spend without a subscription, and pricing across images, Kling and Omni now reflects real production value.

This release brings a faster Vision canvas, a smarter cast, and a new way to earn alongside Plumar.

Vision gets sharper hands: Seedance 2.5 now runs a full 30 seconds, Shot Director speaks through Opus 5, and your most-used nodes rise to the top of the palette. Face casting is calmer too — verified performers are remembered, single frames can be pruned, and the Human ID walkthrough now guides newcomers in two languages.

We also opened the Plumar Referral Program — three tiers, transparent payouts, and a live dashboard that shows who you brought in and what they earned you, month by month.

Under the hood, the studio feels lighter across the board: menus glide, previews load instantly, and every screen respects your currency and your time.

This release brings longer shots, a smarter Shot Director, and a cleaner way to work with the people in your films.

Seedance 2.5 is now live inside Vision, extending your single-shot runtime to a full 30 seconds — enough room for a real performance to breathe. The engine badge always shows which model is driving your shot, and the model switcher finally does exactly what it promises.

Your Shot Director now runs on Opus 5, so scene breakdowns and rewrites land with more nuance and intent. Instead of regenerating a whole shot when a line is almost right, you can now refine the text directly and keep the take you love.

We also polished the people side of Vision: confirmed client faces, a step-by-step Human ID guide in two languages, cleaner frame previews, and a glass-panel dossier that reads while your director is still thinking.

Your Storyboard now thinks like an editor — and your films now speak in any voice, in any language.

The Storyboard has grown up. Instead of stretching a single take, it now cuts on the beat — planning a scene as a sequence of 15-second shots, each with its own framing and intent. Feed it a raw idea or a finished concept, and it lays out the coverage a director would actually shoot.

We've also laid the foundation of our in-house dubbing: speech is transcribed with precise timecodes, split into clean lines, and re-voiced so the new track lands exactly where the original did. Alongside it, ElevenLabs arrives as a second voice engine, with auto-dubbing and cinematic sound effects built in.

Add the shared Depot library, smarter Deck cuts on real edit points, and the pinned REF that keeps your character consistent across every shot — and this is the most complete Plumar yet.

Your Deck just grew up — a shared library, smarter cuts, and no more déjà vu.

We've opened the Deck into a shared library, so every shot, look and selection lives in one place you can browse, revisit and build from. Curation, quality and your wallet timing all feel calmer and more deliberate.

Splices now break on the real seams of your footage instead of dividing time evenly — cuts land where the story actually turns. And a new déjà-vu guard quietly keeps near-duplicate frames from crowding your shelves.

Small polish, too: shelf rows are cleaner without the hotkey clutter, and every frame now carries its caption in both languages.

Every film starts as a folder of screenshots nobody can search. Deck is that folder, alive: drop a frame in and it comes back read — shot size, light, optics, mood, palette — and one click away from becoming your next shot.

Throw frames in any way you like: drag them onto the canvas, paste from the clipboard, or drop a link. The moment a frame lands, the AI cataloguer looks at it the way a DOP would — shot size, lighting, optics, mood, genre, interior or exterior, time of day — writes a one-line caption and pulls a five-colour palette straight out of the image. No waiting, no button to press: the card arrives already read.

Drop a collage and it does not stay a collage. A contact sheet, a storyboard strip, four stills stacked in a column — Deck recognises the montage, cuts it into separate frames and catalogues every one of them on its own. One link in, four usable references out.

Then find them in your own words. Search speaks your language, not the tag list: ask for a close-up, fog or night and it knows which English tags you meant. Filter by what your library actually contains, open any frame full-screen with its metadata, copy a colour out of the palette, and let similar shots pull the neighbours that share its language.

And because Deck lives on the canvas, a reference never stays a picture. Select the frames you want and send them to the canvas as image nodes — or extract the look: Plumar reads only tone, light and optics out of the frame and hands you a prompt, without dragging the setting along. Your taste, finally searchable. Press R.

Fifteen seconds was a shot. A minute is a scene. Today the Seedance duration slider stretches to 60 seconds — one prompt in, one continuous minute of cinema out.

Pull the slider past fifteen — up to a full minute. Behind the curtain, the AI director reads your prompt and plans the minute the way a filmmaker would: a sequence of shots, each with its own camera move and its own beat, all serving one scene. Then it films them one by one and delivers one seamless MP4 — you never touch the seams.

Consistency is the whole craft here. Every shot hands its final frame to the next one as a living reference — the geometry of the room continues even across the cuts. And the plan carries a set lock: the same set, props and light are written into every shot, so nothing drifts. One face, one wardrobe, one world — held for the entire minute.

It works right on the Seedance node, up to 4K. The node shows progress shot by shot while the sequence is being filmed, and billing stays fair: you pay for delivered seconds — a shot that never arrives is never charged.

This is the standard we build toward: a prompt is not a clip — it is a scene. Write one line. Direct. Take the minute.

A film score sets the mood. A song becomes the scene. Today Plumar connects Suno V5 — full songs with real vocals and lyrics, born right on the same canvas as your frames.

The Soundtrack node now carries two engines: Lyria 3 for cinematic score, and Suno V5 / V4.5 for complete songs — up to eight minutes, with a voice you choose: male, female or auto. Three ways to work: pure instrumental; your own lyrics — verse and chorus tags supported; or just describe the song and Suno writes the words itself.

Every run returns two takes — both playing right on the node. Compare, pick the lead take with one click, and that is what travels down your chain; both takes land in your Assets, stored permanently in your library.

And the Sound Director — our composer brain — has learned to write for a singer. Give it the idea in plain words, and it delivers a full song brief: a genre-true style formula, a title, structured lyrics in the language of your intent — the node picks it apart automatically. There is a one-click Song (Suno) chain in the gallery: idea → composer → track, already wired.

This is the standard we build toward: if a scene needs a song, the song is born where the scene is. Write. Direct. Now — sing.

Every director knows the moment: a node turns red, the render stops, and the question is always the same — why? Today Plumar ships Plumar Support — a built-in assistant that lives right on the Vision canvas and answers about your project, not in general words.

Tap the lifebuoy in the bottom bar and a chat docks to the right of your canvas. Ask it anything about the platform — where the wallet is, how to renew, where to write your curator — and it doesn't just explain: it takes you straight to the right section. Ask about money and it reads your real balance and subscription. It never touches your funds — reading only.

The part we're proud of: error diagnosis. Ask "why is my node red" and it walks UP your chain like an engineer — finds the real root (an empty Prompt three nodes above, a failed ancestor, a missing connection) and highlights the culprit right on the canvas. It also tidies your layout on request (fully undoable) and suggests what to connect next — only nodes that actually fit.

Here is the engineering trick: 80–90% of answers need no AI at all — they come from the platform itself, instantly and for free. That's why Support is on for every student, not a premium add-on. Only open questions go to the Gemini brain, and Academy questions get answered from the real knowledge base.

Every conversation is saved to your account, so the team sees the full story when you need a human. Before release we ran an adversarial bug-review and fixed every finding. A film set runs on a good crew. Now your canvas has one.

Every AI image has a tell: waxy skin, dead eyes, that faint plastic sheen that whispers rendered. Today Plumar connects Seedream 5.0 Pro — ByteDance's flagship image model and the most realistic engine we have ever put on the canvas.

What sets it apart is the surface of life itself: skin with real pores and fine wrinkles, individual eyelashes and stray hairs, fabric weave you can almost feel, light that falls and scatters the way it does on a film set — not the airbrushed averages most models settle for. Faces hold their character instead of collapsing into a generic beauty mask.

It goes up to 4K — enough for a cinema master frame, a poster, a print. And it is not only a generator: feed it reference images and it edits — keeping your character, your location, your light while changing exactly what you ask. The frames it produces carry straight into video: a more real keyframe makes a more real shot.

Using it takes one click: drop an Image node, open the panel, pick Seedream 5 Pro. It slots into every chain and workflow you already have — Cast Cards, Coverage, Scene Shots, Realism — the whole pipeline simply gets a sharper eye.

This is the standard we hold every tool to: if a frame doesn't look alive, it isn't finished. Pick the model. Direct.

The biggest barrier in filmmaking was never talent — it was knowing what to do next. Today Plumar removes it twice: every node now teaches itself, step by step, and 45 ready workflows put a whole production playbook one click away.

Drop any node on the canvas and it walks you through like a first AC on set: next → next → done. One idea per step — what this node is, which model to pick and when, what the formats mean, where the wires go, what it costs. No manuals, no guesswork: thirty nodes, thirty built-in walkthroughs, each written for someone who has never touched a node editor. Dismiss it once for a node you've mastered — it stays quiet for that node only.

The Chains library grew from a starter pack into a studio playbook: 45 wired, validated workflows across seven categories — character, scene, sound, advertising, screenplay, VFX, depth. A frame with a consistent face. A dialogue scene with correct eyelines. A vertical reel with the score already inside the file. A full series pipeline where the show bible keeps your hero's scar on the correct cheek across episodes. Every card shows the live map of what lands on your canvas; every chain arrives connected, valid and pre-filled with working prompts.

And the toolset those chains draw from got deeper. DepthLab reads the depth of any frame — isolate a slice of distance as a clean matte, swap a background without touching the actor, click anywhere to set focus with real f-stops. A VFX floor arrived alongside: mux the score into the mp4, lay cinema looks over finished clips, master one shot for cinema, YouTube and Shorts, widen frames to anamorphic, relight day into night, empty a set of people. The writers' room got a spine too: Show Bible → Script → Shot Breakdown — one click scatters an episode onto the canvas as ready shot prompts.

This is the thesis behind Plumar: the craft should live in the tool, not stand between you and the film. Pick a chain. Follow the steps. Direct.

Until now AI video had one unbreakable rule: no real faces. We built the path around it — officially. Upload a photo, and Seedance 2.0 shoots the scene with your face.

The Video node gets a new engine: Seedance 2.0 · Faces — the only route in Plumar that officially accepts real human faces as references. Not a workaround, not a jailbreak: every photo travels through a dedicated asset pipeline built exactly for this. Attach a portrait to the CAST input — and the person in the frame is recognisably, consistently them, shot after shot.

It's the same Seedance 2.0 cinema engine — in two tempers: Quality up to 1080p for hero shots, Fast up to 720p for iterating. Any duration from 4 to 15 seconds, up to 9 references in one shot — faces, frames, even video and audio — and native sound in the take. Prices didn't move.

What it unlocks is bigger than a feature: you become castable. The director's own face in a teaser. A founder presenting inside their product film. A client's face in the commercial you're pitching tomorrow. Where AI video used to swap in a stranger, Plumar now keeps the person — because in cinema, the face is the story.

Open the Video node, switch the provider to Faces, connect a portrait to CAST — and shoot yourself into the scene.

The hardest part of a node canvas is the blank page: which nodes, in what order, wired how? Chains answers it. Pick the task — a ready, fully-wired workflow lands on your canvas.

Open Chains from the palette (or straight from an empty canvas) and you're looking at a gallery of production recipes: Character frame. Scene coverage. Shot from a frame. Product ad. Music clip. Video dubbing. The full shot — frame to video to score. Each card shows a live map of its chain — the actual nodes and wires you're about to get.

One click and the whole chain lands on the canvas — already connected, already valid. Prompts come pre-filled with working examples, a short note explains the flow, and the camera glides to your new setup. Replace the example text with your story and hit generate. No guessing which output feeds which input — the wiring is the craft, and it's done.

Under the hood each chain is a living recipe, not a frozen snapshot: nodes are born with the current defaults and models at the moment of insertion, and every connection passes the same validation as hand-wiring. Chains live in the database — new ones will keep appearing without waiting for an app update.

This is how Plumar thinks about the canvas: the craft of wiring should never stand between you and the film. Pick the task. Fill in your story. Direct.

One frame used to cost you the wait. Now, in that same time, Plumar lays up to eight frames at once onto the canvas — and gives them a new eye.

A new image model joins the canvas: GPT Image 2 by OpenAI. It sits alongside Nano Banana 2 and Nano Banana Pro — pick it on any Image node, at 512 / 1K / 2K / 4K. Each model has its own eye: Nano Banana holds photoreal, cinematic texture, while GPT Image 2 is stronger on composition, typography and stylization. One frame for mood, another for a layout with text in the frame. Switch the model right on the node, no rebuilding the graph. One moment, two cinematographers.

The Image node now has a multiplier: ×2, ×4, ×6, ×8. One click and Plumar renders the whole set at once — same prompt, same references, in parallel. Not a queue of eight runs, but a single pass that lands as a clean grid on the canvas. Six angles on a moment. Eight looks for a key frame. A full row of storyboard — in one go.

Every frame is its own compact tile: just the image, an input on the left, an output on the right. Drag the one you like into a Video node, into Edit, or feed it as a reference — the connections behave like any other node. Need the controls — the "expand" button turns the tile into a full Image node. The graph stays clean: eight variants don't become eight heavy panels.

You pay only for frames that land. A frame fails — it's free, one click to retry, the rest of the grid untouched. Before it starts, Plumar checks you have frames for the whole set — no surprises mid-render.

Storyboarding stops being the bottleneck. You see a moment from six angles and choose the frame with a director's eye, instead of guessing from a single try. Storyboarding becomes a look again, not a queue. Open any project, pick a model, set ×6 — and see the scene from every side at once.

Getting great results out of AI used to demand great prompts — naming the lens, the light, the tempo, the exact edit. That's a craft most people don't have. Plumar removes it: describe what you want in ordinary words, and a director brain turns it into a full professional prompt built for the model.

Every AI tool has the same catch — it only gives you what you asked for, and asking well is a skill. Type "make it cinematic" and you get nothing back. A prompt that actually works names the focal length, the colour of the light, the film stock, the beats per minute, the key, the precise change. Learn all of that, or live with mediocre. Until now, that was the price of admission.

Plumar pays it for you. Anywhere on the canvas, write a plain wish and press one button — a director brain, trained on how each model actually behaves, rewrites it into a precise, professional prompt. The jargon stays under the hood; you just get the shot, the score, the edit you pictured in your head.

See it work. "Replace the background with a cozy cafe" becomes a clean, model-ready edit that keeps your character and their motion untouched. "Tense strings under a chase" becomes a full musical brief — genre, tempo, key, instrumentation, structure. "A wide dawn shot on the cliff" becomes a cinematographer's prompt with lens, lighting and camera movement.

Three directors, one language — yours. Shot Director composes the frame: framing, lens, light, camera move. Sound Director writes the score: genre, tempo, key, instruments, structure. And the new Edit Director handles video edits: swap a background, change the weather, restyle a whole scene. Each one has memorised its model's rulebook so you never have to open it.

It's honest with you, too. Ask for something a model genuinely can't do — a 4K upscale, an edit it can't make — and the director tells you upfront, instead of quietly burning a render and your frames on a dead end.

That's the whole idea behind Plumar: you direct in plain words, and the machine speaks fluent cinema. No prompt-engineering, no manuals, no guesswork — just the result you actually meant.

Plumar's canvas gets sound. A new Sound Director brain turns a plain wish into a professional prompt — and Lyria renders the music.

Drop a Soundtrack node on the canvas and type the mood — "tense strings under a chase", "warm piano for the reunion" — and Google Lyria 3 composes an original track. A 30-second cue or a full song, instrumental or with vocals.

Don't know the musical words? The new Sound Director — a Claude Opus brain, the same engine behind Shot Director — turns your plain wish into a full professional brief: genre, tempo, key, instruments, structure, length. You speak human; it speaks music.

It even reads your frames: feed a shot into the node and the score is written to match the light and mood on screen.

A new engine joins the studio. Google’s Gemini Omni now renders your scenes with native sound — and soon you’ll re-direct a finished shot just by telling it what to change.

Omni Flash is wired straight into the Video node — no middlemen. Pick Gemini Omni as your provider and get 3–10 second shots with sound designed by the model itself: ambience, foley, even dialogue.

Feed it up to six reference images — your cast, your location — and it keeps them recognizably yours, frame after frame.

And this is only phase one. Omni understands spoken direction: “make it dusk”, “slow the camera”, “lose the crowd”. Conversational editing arrives in Plumar next — same faces, same world, new take.

A new kind of studio just opened. Plumar turns a logline and a cast into finished cinematic video — you write, cast, shoot and edit, end to end.

Today Plumar goes live. Direct entire scenes with 99% consistency — the same faces, wardrobe, locations and lighting holding from the first frame to the last.

No prompts to wrestle, no setup. Point the Director at your story and it plans the shots, casts your characters, and renders them with a locked cinematic look.

Early access is rolling out in waves — leave your email and we'll bring you in.