TRAG Studio is organized around three categories — Image, Video and Audio — reachable through the tabs at the top of the screen. Each category contains workflows: ready-to-use creative studios.

Who it's for

Throughout this guide you'll see screenshots of the app: the bright-green boxes and arrows in them point at the element being discussed. Click any image to enlarge it.

The three offers: Free, Pro and Trag Muse

TRAG comes in two editions, plus an optional automation layer:

Two possible paths, your pick:

• Start free, go Pro later. Download TRAG Studio Free from the website, create images at will… then, when video or audio calls, buy Pro on the web portal. The key is attached to your profile: the app unlocks the workflows and offers to download the missing models.
• Buy Pro right away. The purchase places the key in your profile; on first sign-in, onboarding installs the complete experience from the start.

Inside the app, reserved workflows carry a PRO badge (e.g. “Magic Edit Pro”) — that badge is the visual marker of this matrix. The full workflow-by-workflow matrix is recapped in the FAQ.

02Installation

Installation uses a classic Windows wizard — a few clicks on “Next” and you're done. Here is the exact journey, screen by screen.

1. Run the TRAG_Studio_V1.0_Setup file downloaded from the TRAG website (about 1 GB).
   Windows asks for a security confirmation: accept to continue.
2. Extract the installer. A small self-extracting window opens, pointed at your Downloads folder. Click Extract and let it finish.

3. Choose your options in TRAG Studio Setup. The installer opens on a single screen: the install folder is pre-filled with your user folder (…\AppData\Local\Programs\TRAG Studio — no administrator rights needed), with two boxes already ticked: desktop shortcut and firewall rule. Leave them on, then click Install TRAG Studio →.
4. Let the installer work. While a slideshow presents the studio, the progress bar tells you exactly what is happening — including the step where TRAG checks which CUDA version your graphics driver supports and fetches the matching build. Nothing to do but wait.

5. Finish and launch. “Installation terminée” offers to delete the temporary setup file (about 2.5 GB — tick it to free the space) and a big Launch TRAG Studio → button. On first start, TRAG greets you with the sign-in screen — on to the next section!

03Sign-in & first-time setup

On first launch, TRAG walks you through a short setup: sign-in, hardware detection, workflow picks, output folder. Five minutes, tops (model downloads aside).

Sign in

The “Welcome back” screen offers two doors: Continue with Google, or e-mail and password. With Google, the sign-in finishes in your browser — TRAG waits, then continues automatically the moment you are done. No account yet? “Create one” takes a moment.

The onboarding welcome

TRAG introduces itself in one line — “Local generative AI, made for everyone.” — and restates the promise: a simple interface running powerful open-source engines on your machine, with no shady downloads and no steep learning curve. Click Get started →.

Hardware detection and the Studio Index

To suggest the right defaults, TRAG inspects your machine. On “Tell us about your machine”, click Detect (recommended): in a second, the app identifies your graphics card and processor, scores each one, then computes your Studio Index — the overall rating that calibrates the app to your PC's muscle.

Prefer to stay in control? “Set up manually instead” unfolds three pickers — graphics card, processor, memory — and previews the expected performance (Image / Video / Audio) live. The “I don't know my hardware” option picks cautious defaults for you.

Minimum configuration

TRAG adapts to your machine — and the floor is lower than you'd think:

Below that, the Studio Index simply picks humbler settings; above it, every extra gigabyte of VRAM buys speed and resolution.

What your machine actually changes

Beyond the floor above, three things decide how TRAG feels day to day: free disk space, video memory, and whether an NVIDIA card is present at all.

Disk space — the real first wall

The models install inside the application folder, pack by pack. The exact weights:

• To install and make images: plan for 50 GB free.
• To have everything, Pro included: plan for 120 GB free.
• On a laptop with a small system SSD, this — not the graphics card — is usually the real blocking point.
• Nothing forces you to take it all: tick what you want, add the rest later without losing anything.
• Your creations go to the output folder you pick at the next step — somewhere else, and that adds up over time too.

VRAM — what it actually buys

• VRAM decides video, not image. Image runs comfortably from 4 GB; video is what needs headroom.
• 4 GB — images without worry. Video: short clips at modest definition.
• 6 GB — comfortable video in short clips. Measured on a tester's laptop RTX 3060: 4.9 GB used out of 6, no incident.
• 8 GB and up — longer durations, higher definitions, bigger batches.

And as described in the interface tour, only one engine stays warm at a time — a choice made to spare your card, not a limitation.

No NVIDIA card? TRAG still runs

Without a usable GPU, TRAG falls back to the processor: the footer indicator announces it in amber, and generation works — just much more slowly. Fine for trying things out, not for producing. Better said here than mistaken for a breakdown.

Pick your workflows — and let the screen check your disk

The models screen lists the packs — Image, Video, Audio — with their real download sizes; the core engines download automatically. As the screen puts it: stills are free, motion & sound are Pro. Every pack can be downloaded — the PRO stamp simply tells you what your plan can run.

Three details make this screen one you can trust:

• It re-reads your disk every time it opens. The banner at the top announces it — “Checking what is already on this machine…”, then the verdict: “Disk checked · 14 of 18 model files verified · 9.4 GB left to download” — with a Re-check button.
• It knows three states, not two. A complete, verified pack wears a green Already downloaded stamp (no checkbox — a discreet Download again remains). An interrupted file shows amber with Check needed and a Redownload button. Anything absent keeps its checkbox and its exact remaining weight.
• Closing the window cancels nothing. The download card lives on; collapsed, it keeps the percentage and the bar visible — “Keep working — this continues in the background.”

And you get to work early: as soon as the essentials are in (about 14.5 GB), the studio opens and the rest keeps downloading in the background. You can add or remove packs anytime via Settings › Hardware & Models — models already on disk are never fetched twice.

Choose the output folder

“Where should your creations be saved?”: pick the folder where your images, videos and tracks will land. Click Choose…, select (or create) a folder in Windows Explorer, then confirm with Continue →. You can change it anytime in Settings.

Home stretch

TRAG shows three things worth remembering — the Main chat bar and its ✦ magic wand (your prompt upgraded in one click), Add Multiple Jobs (run a whole batch in one go) and the Media Gallery (all your creations, ready for reuse). Then “You're all set” confirms your engines are installed: click Let's Create →.

The guided tour (8 steps)

On first open, a short guided tour spotlights the key areas one by one (mode picker, workflows, settings, prompt…). Follow along with Next →, or dismiss it with Skip — you can relaunch it anytime via “Take the tour” in the sidebar.

04The interface at a glance

All of TRAG fits on one screen, built so you never get lost. Here's the map.

On startup, TRAG preloads its models (“Warming up the models”), then drops you into the Create space:

• The sidebar (left) — Search (the command palette), Create (the studio), Media Gallery (your creations), Creation Flow (the render queue) and Cinema Flow (a preview of the film timeline — coming soon), then at the bottom: Light mode, Settings, Help, Toggle Fullscreen and your profile.
• The Image / Video / Audio tabs (top center) — the creation category. Each category then shows its workflow sub-tabs (Text to Image, Image to Image, Magic Edit, Blend Images…).
• The viewport (center) — the stage where your result appears, compares, replays.
• The settings panel (right) — the options of the current workflow: Dimensions, Style, Quality… “More settings” unfolds the advanced ones.
• The Main chat bar (bottom) — describe your idea, then hit the big action button (Generate Image, Animate Image, Generate Music…). The ✦ magic wand upgrades your prompt in one click, the mic takes voice dictation, and an example suggestion sits underneath for inspiration.
• The Warm up button (top left) — preheats the active category's engine so the first render starts without delay.
• The status bar (very bottom) — the health of the local services: “AI engine connected” (the generation engine, with GPU activity) and “AI Companion connected” (the prompt assistant, on CPU).

The Warm up button and engine status

Each category has its own engine, and TRAG keeps only one warm at a time to be kind to your graphics card. The badge at the top left tells the whole story:

A fourth state exists: when no GPU is usable, the footer indicator turns amber and says so plainly — “AI engine connected (CPU — no GPU detected)”. That is the processor fallback described in section 03: everything works, just much more slowly.

Light, dark… or Creator

The first button at the bottom of the sidebar switches the mood: a clean light mode, a deep dark mode, and Creator mode — a matte black made for distraction-free creating, with optional accent flavors (like “Obsidian Orange”). A small toast confirms each appearance change.

The search palette

The sidebar's Search button opens a command palette: type to filter screens, workflows and actions (“Go to Media Gallery”, “Text to Image”, “Open Settings”…) and confirm with Enter. Esc closes it.

05Create images

The Image tab gathers four workflows: create from text, transform a picture, retouch one exact spot, or merge several images into one.

Text to Image — create from text

“Create a brand-new image from text alone.” The starting workflow, and the most direct:

1. Image tab → Text to Image sub-tab.
2. Pick your settings on the right: Dimensions (Portrait, Square, Landscape, Poster or Custom) and Style (Auto, Photo, Cinematic, Illustration, 3D, Anime, Watercolor, Oil Painting, Pixel Art, Comic, Fantasy).
   “More settings” adds Custom settings, a Creativity dial and a negative prompt.
3. Describe your image in the bottom bar — subject, style, lighting, mood.
4. Click Generate Image →. The result lands in the viewport and joins the Media Gallery automatically.

After generating, two buttons appear on the result: ✦ Re-use prompt (start again from the same text) and ✦ Use as source (send the image to another workflow). The Recent Generations strip under the viewport keeps your latest tries one click away.

Image to Image — transform a picture

“Restyle, enhance, mix or upscale an existing image.” Hand over a starting image, describe the change:

1. Image to Image sub-tab, then load your picture via Source image → Click to choose (or tick “Iterate from last result” to build on the previous render).
2. Set the Style, the Quality and the Upscale factor (1x / 1.5x / 2x — great for a crisp enlargement).
3. Describe the transformation — “restyle, enhance, upscale…” — then click Transform Image →.
4. Compare. The viewport shows a Before / After slider: drag it to sweep between original and result.

Magic Edit — repaint one exact spot Pro

“Paint a region and replace just that part — add, remove or swap objects.” Deliciously simple: you paint the area to change, you say what goes there, TRAG touches nothing else.

1. Magic Edit sub-tab, then load the image to retouch (Image to edit → Click to choose).
2. Pick your mask in the right panel: Brush (freehand) or Auto-detect. Adjust the brush's Opacity and Hardness if needed — Brush size and Blending strength live under “More settings”.
3. Paint the area to change right on the image — it gets a highlight veil. “Clear canvas” wipes the mask, “Reuse mask” brings back the previous one.
4. Describe the change in the bottom bar (add, remove, replace…) then click Apply Edit →. The Before / After comparator then lets you check the retouch.

Blend Images — merge several pictures Pro

“Combine multiple images into one — subject + style, or several references.” A subject from here, a style from there: Blend marries two images (plus an optional third).

1. Blend Images sub-tab, then load Image A and Image B (“+ Add another” for an optional C).
2. Pick a Style if you want to steer the look.
3. Describe how the images combine — who does what, in which scene — then click Blend →.

06Create videos Pro

The Video tab creates a clip from a prompt alone, animates your images, bridges two moments with fluid motion, or makes a face speak. All four video workflows belong to the Pro edition.

Text to Video — a clip from a prompt alone

“Create a video entirely from a text description.” No source image needed: describe the scene and the motion, and TRAG films it.

1. Video tab → Text to Video sub-tab.
2. Pick the frame: Format (TikTok, YouTube, Square or Custom), Aspect ratio (9:16, 16:9, 1:1, 4:3), Definition (SD 480p → UHD 2160p) and Duration (5s, 10s, 15s or Custom).
3. Describe the shot — subject, camera move, mood — then click the action button and let the engine film.

Image to Video — animate a still

“Bring a still image to life with motion.” One image + one motion description = a video clip, with synchronized audio and dialogue if you describe them.

1. Video tab → Image to Video sub-tab, then load your image (Source image → Click to choose, or send one from the Media Gallery with “Use as source”).
2. Pick the Format — TikTok (vertical), YouTube (16:9), Square or Custom — and the Duration (5s, 10s, 15s or Custom). Quality and Smoothness (24 / 30 / 60 FPS) sit under “More settings”.
3. Describe the motion — what the subject does, the camera move, the soundscape, even a line to speak.
4. Click Animate Image → and let the engine work: the clip then plays right in the viewport.

Smooth Motion — from one frame to another

“Morph a start frame into an end frame with fluid motion.” Provide the starting picture and the ending picture: TRAG invents the fluid motion that joins them.

1. Smooth Motion sub-tab, then load the Start frame and the End frame.
2. Choose the frame and fluidity: Format, Duration — and Smoothness (24 / 30 / 60 FPS) under “More settings”.
3. Describe the motion from first to last frame, then click Generate Motion →.

Lip Sync — make a face speak

“Drive a face video or image with an audio track so the lips match.” A face (photo or video) + an audio track = perfectly synced lips.

1. Lip Sync sub-tab, then load the Face video/image and the Audio track.
   No audio file handy? Record from mic captures your voice on the spot.
2. Pick the frame: Format and Duration — Quality and Smoothness under “More settings”.
3. Add an optional prompt to guide the render, then click Sync Lips →.

07Create audio Pro

The Audio tab composes whole tracks — with or without lyrics — remixes your music and regenerates one exact passage. All three audio workflows belong to the Pro edition.

Text to Music — compose a track

“Compose music from a description.” Describe the track, set the mood… and write lyrics if you feel like it:

1. Audio tab → Text to Music sub-tab. The top button becomes “Warm up ACE-Step 1.5 engine”: one click to preheat.
2. Write your lyrics (optional) in the central Lyrics editor: structure them with the Verse, Chorus and Bridge buttons, pick the language (EN, FR, ES, DE, IT, PT, JA, ZH, KO, RU…). Leave empty for an instrumental.
3. Shape the music on the right: Style (Pop, Rap, Afro, Rock, Electronic, Cinematic), Mood (Happy, Epic, Chill… sixteen moods), Key & Scale (the little piano keyboard!), BPM and Duration (30s, 1m, 2m or Custom).
4. Describe the track in the bottom bar, then click Generate Music →. The track shows up in a player, ready to listen — and filed in the Media Gallery.

Once the track is generated, the player's ✦ Use as source button sends it one click away to Remix this track, Section Edit or Lip Sync (audio):

Remix — restyle a track

“Transform an existing audio track into a new style.” Your song, a new outfit:

1. Remix sub-tab, then load the track (Source audio → Click to choose).
2. Pick the Style and the Mood, and dose Keep Original (under “More settings”): the higher the percentage, the more the original shows through.
3. Describe the new direction — genre, vibe, tempo — then click Remix Track →.

Section Edit — regenerate a passage

“Replace or regenerate a chosen section of a track.” Chorus falling flat? Regenerate just that, leaving the rest untouched:

1. Section Edit sub-tab, then load the track.
2. Mark the passage to rework directly on the track in the viewport.
3. Choose Match: Keep key & tempo to stay in step with the rest of the track, or Free to let go.
4. Describe the replacement then click Repaint Section →.

08Creation Flow & batch jobs

The Creation Flow is your render control tower: the live queue, the history of everything you've made, and batch launching.

Follow the render queue

Open Creation Flow from the sidebar (the badge on the icon counts waiting jobs). Each row shows its state — Generating with percentage and time left, Waiting with an estimate, then Completed — plus its Open / Cancel buttons. Top right: History, Add Multiple Jobs, and while running, Stop, Pause and Stop all.

The creation history

The History button opens the Creation History: session stats (creations per workflow) and, for each entry, Use prompt (reuse the text), Save preset (memorize the settings) and Open. Export saves the history, Clear empties it.

Add Multiple Jobs — launch a whole batch

Feel like generating ten images and five tracks over lunch? Add Multiple Jobs runs a whole list of creations, automatically:

1. Copy the “matrix” of the category you want — Copy Image, Copy Video or Copy Audio. It's a text template that teaches an AI (ChatGPT, Claude…) how to write your job list.
2. Have your list written: paste the matrix into your favorite AI along with your ideas, get back a .md file (one job per line, a # heading per category).
3. Drop the .md onto the “Drag result here” zone: TRAG reads it and announces “X jobs ready”. If your jobs use files (images, audio…), also drop a .zip with those assets onto the second zone.
4. Choose the output folder then click Generate: the jobs roll through the Creation Flow on their own, and every render joins the Media Gallery.

09Media Gallery

Every creation — image, video, track — is automatically filed in the Media Gallery. It's your library: find, replay, reuse.

Browse and filter

At the top of the gallery: filters by type (All / Image / Video / Audio), by origin (All / Mine / Agent sessions — handy to isolate what an agent created), sorting (Newest first…), grid size and Refresh.

A media's file card: Item Details

The Item Details panel shows everything worth knowing: preview (or player), file name, type, size, date, source workflow — even “Request from: AI agent” when a creation came from an agent. Further down, the full prompt with three buttons: Copy, Send to chat (send it back to the Main chat bar to rework it) and Save as preset. Finally, Save As… exports the file wherever you like, and Delete removes it.

Reuse a creation: Use as source

The ✦ Use as source button is the bridge between your workflows: it sends the selected media off as the starting point of a new creation. For an image: Image to Image, Magic Edit, Blend Images (A or B) or Image to Video. For a video: Smooth Motion or Lip Sync (Face). For a track: Remix, Section Edit or Lip Sync (audio).

10Settings

The sidebar's Settings button opens the preferences, organized by topic: Connection, Storage, Hardware & Models, Preferences and Language.

Folders, hardware and models

• Storage — the Gallery folder (where your library lives) and the Generated files folder (where new creations are saved): two Browse… buttons to change them.
• Hardware & Models — the current hardware diagnosis, Re-scan hardware to run detection again, manual selection, and Installed models & workflows to add or remove downloaded engines.
• Preferences — interface sounds, the startup tutorial, custom shortcuts (Alt+F toggles fullscreen) and display scale.
• Language — the interface language.

At the bottom of the window: Reset restores defaults, Save changes keeps yours.

Your account, your license

A click on your profile, at the bottom of the sidebar, opens the account panel: Manage account & billing, your plan, the app version with its Check for updates button, your communication preferences — and Report a problem, the direct line when something goes wrong. This is where you check that your Pro license is active.

How TRAG keeps itself up to date

• TRAG checks silently at startup, then every 6 hours. If there is nothing new, you see nothing.
• When a version exists, a discreet banner appears at the bottom left. Never blocking — keep working; Later snoozes it for 24 hours.
• One click on Update now: the app closes itself, updates, and relaunches. A patch weighs a few megabytes — not the installer's gigabytes.
• Models already downloaded are never re-downloaded by an update.
• Prefer to check yourself? Account panel → Check for updates.

Trag Muse & Agent Gateway Muse

Trag Muse turns your workstation into a connected production studio: your agents and your APIs drive TRAG from first idea to final master — while your files, your models and your rendering stay on your own machine.

The plumbing behind it is the Agent Gateway, in Settings › Connection — the fuchsia-trimmed section. It stays off until Muse opens; here is what the setup will look like then:

1. Turn on the Agent gateway (switch to On — the status reads “● Running”).
2. Note the local address under “Reachable at” (with its copy button) — that's what your agent will use. The Local network option also exposes it to your Wi-Fi if you drive TRAG from another device.
3. Download the configuration with Save mcp.json, then follow the on-screen hints to plug it into your agent (Claude, Codex…).
4. Leave the gateway On: your agent can now create in TRAG — its works are tagged “Request from: AI agent” in the gallery, filterable via “Agent sessions”.

Clicked “Cancel” on the firewall alert? Nothing is broken

Everything TRAG does on its own — talking to its engines on your machine, downloading models, checking for updates — needs no firewall permission at all. The one and only case that does: being driven from another device on your network. So if you refused the Windows alert, neither downloads nor generation nor updates were harmed. To grant the permission after the fact:

Path 1 — through Windows settings (no command line)

1. Open Start, type firewall and choose Windows Defender Firewall.
2. Click Allow an app or feature through Windows Defender Firewall.
3. Click Change settings (asks for administrator rights).
4. Find TRAG Studio in the list.
   If it is there: tick Private. Do not tick “Public”. If it is not: Allow another app… → Browse… → select TRAG.exe in the installation folder → Add, then tick Private.
5. Confirm with OK and restart TRAG.

Path 2 — one command (for those at ease): open a command prompt as administrator, and adapt the path if needed:

netsh advfirewall firewall add rule name="TRAG Studio" dir=in action=allow ^
  program="%LOCALAPPDATA%\Programs\TRAG Studio\TRAG.exe" enable=yes profile=private

Two nearby cases worth knowing

• Third-party antivirus. Some block the Windows firewall prompt, or the downloader itself. If TRAG cannot download models while your connection works fine, temporarily pause the antivirus, or add the installation folder to its exclusions.
• Company or school network. A proxy or filter can block download.trag.studio. The symptom: models never start downloading while the rest of the internet works. Try a home connection or a mobile hotspot, or ask the network administrator to allow that domain.

11Tips for better results

A few habits that change everything, from first prompt to final render.

The Main chat bar, your best ally

Everything goes through the Main chat bar, the single bar at the bottom of the screen. Describe your idea, then let its tools work: the ✦ magic wand rewrites and strengthens your prompt in one click (sharper subject, lighting, mood), the mic takes voice dictation, the paperclip attaches a reference, and the ✦ suggestions underneath show the level of detail that works. From the gallery, “Send to chat” drops an old prompt back into the bar to rework it into something new.

12FAQ — Frequently asked questions

Short answers to the questions we hear most.