Settings & Publishing

1. Project Settings

Project settings affect the selected project and can override the global defaults for that project.

Guidance (CFG) controls how strictly the AI follows your prompts. Video guidance (Video CFG) sets how closely the video matches the scene description, and Audio guidance (Audio CFG) sets how closely the generated audio matches the sound description. Higher values force stricter adherence to the prompt; lower values allow more creative freedom.

2. Global Settings

Settings holds your account and billing, the defaults new projects start from, and the services the desktop app connects to. Open it from the user/profile area in the left sidebar.

The Settings navigation lists these tabs, in this order:

When the list is taller than the window it scrolls, with a fade and a chevron (More settings) at the bottom.

General

Usage

Usage shows the credits included with your plan, credits used and remaining, renewal or cancellation information, and usage over time. Depending on where the subscription is billed, it also offers plan-management, upgrade, or pending-change actions.

Under the plan name a line says where the plan is billed:

Free daily credits and the free trial are described under "Credits, Plans, Free Trial, and Daily Credits".

Generation Job History

At the bottom, click Load latest generation jobs to open your account-wide Aificient Cloud video render history. It includes jobs from all projects and lists the newest first. Each job can show:

• Content and scene name.
• Status: Queued, Running, Completed, Failed, Cancelled, or Unknown.
• Quality tier and resolution.
• Requested and finished times.
• Progress percentage plus the current stage or message while the job is queued or running.
• The settled credit price of a completed job, shown with the credit icon (for example 3.4 credits).

Note: A price appears only once billing has successfully settled a completed render. Queued, running, failed, cancelled and unknown jobs do not show one, and a completed job may omit it while billing is not settled.

Use Previous and Next to move through longer histories. The page refreshes automatically while it contains queued or running jobs and stops when no active jobs remain. Use the refresh control to check manually, or Hide to collapse the history.

Storage

Storage shows how much cloud storage your projects and assets use, your plan limit, the percentage consumed, and the remaining space. Warning and critical states explain when storage is close to or over the allowance.

If you need more room, remove unneeded project data or use Upgrade your plan. Storage-limit errors also appear in a banner at the top of the app, with shortcuts to manage storage or view plans.

Notifications

Desktop alerts can be switched on independently for:

• A scene video finishing.
• A final stitched project becoming ready.
• A GPU render queue becoming empty.
• A local or Vast.ai GPU finishing setup and becoming ready.

Social Accounts

This tab connects the social accounts you publish to — the accounts the Publish dialog uses (see "Publishing the Final Video"). Each platform (TikTok, Instagram, YouTube) has its own row. TikTok is supported today; Instagram and YouTube show a Coming soon tag.

Everything here is local to this computer. Connecting an account opens a secure sign-in window, and your session stays on this device — nothing about your accounts is sent to a server.

Note: Refresh metadata checks account details and session state, but it does not replace an expired login — use the amber icon for that.

Video Config

Video Config holds the defaults used when creating new projects:

• Start voice delay.
• End voice delay.
• Transition duration — the default crossfade between scenes of new stories.
• Captions — the switch, a live preview, Position, Font (all 51 fonts, see "Caption Fonts"), Font Size, and the text, highlight, outline and background colors. Turning captions on with no font set selects Bangers.

Note: Changing these global defaults does not alter an existing project. To change captions, timing or other video settings for the selected project, use the right scene sidebar's Settings tab and apply the changes there (see "Captions").

Model Config

Model Config sets the default video-model settings used by new projects:

• Quality (Lite or Pro).
• Social media caption instructions.
• Local runtime download folder.
• Video guidance (Video CFG).
• Audio guidance (Audio CFG).
• Prefetch count.

Social Media Caption Instructions is optional free-form guidance the AI follows when it drafts a publish caption (the Generate button in the Publish dialog). Use it to set a tone, style, call-to-action, emoji use, or other constraints — for example, "casual and punchy, end with a question, no more than two emojis." It applies to every project, and you can leave it blank.

The runtime download folder can only be changed while the local runtime is stopped and is not downloading, setting up, or booting. Use Default to return to the app's standard data folder.

Audio Config

• Speech pace.
• SFX volume — the default level of the video model's own soundtrack under the narration, a 0–100% slider (default 20%). New projects start from this value; each project can override it in its own Settings tab.
• Default narrator voices.
• Automatic voice accent — the Spanish and the English accent an Automatic narrator speaks with.

Default Narrator Voices

The Default narrator voices box (For narrators with no voice of their own, by gender. Automatic picks the library voice for the script's language and accent.) has a Female and a Male row.

How the defaults apply:

• A narrator with no voice of its own is sent with the default for its gender.
• If that voice does not speak the script's language, the service uses its automatic voice instead and says so in a notice on the Narrator card.
• A voice assigned to the narrator itself (on the Narrator card, or in the character editor) always wins over these defaults.

Automatic Voice Accent

The Automatic voice accent box (A narrator on Automatic speaks with a voice of this accent. A voice you choose keeps its own.) has a Spanish and an English row, each with its flag and a dropdown:

Voices

Settings > Voices holds the narrator voices you cloned yourself: Your own narrator voices, cloned from your recordings. Pick them from any voice field; every scene that names one is narrated with it. The heading carries an N/M badge — how many you keep against the number your plan allows.

Each voice is a row with a preview button, its name, trait pills (the gender, and the language with its flag), and a detail line: Cloned from your recordings, plus No preview or Rebuilt on next use when they apply. Rebuilt on next use means the service has to rebuild the voice; nothing is required from you.

Note: While the server narrates with a different service, the section notes that your voices are not available on it right now.

Saved Library Voices

While narration runs on the voice library, a second block, Saved library voices (with its own N/M badge), lists the narrators you bookmarked in the Voice library: Narrators you picked from the voice library. They show up in every voice picker and speak every language.

• Explore opens the Voice library.
• Each row has a preview button, the name, trait pills, a <use case> · <tagline> line, and a remove button (Remove from saved voices (scripts that use it keep it)).
• While the list is empty it reads Nothing saved yet — bookmark voices while exploring.

Cloning a Voice

1. Click New voice. The modal opens with Clone your own voice from a few recordings. It joins your voice library with a preview.
2. Enter a Name.
3. Set Language and Gender. A clone always has a gender, because that is what the voice picker filters on.
4. Add Recordings (1–2 minutes of clean speech) — up to 5 audio files, 20 MB in total, in mp3, wav, m4a, aac, ogg, webm, or flac.
5. Click Create voice · N credits. When it finishes, the modal says "<name>" is ready and the voice can be picked from any voice field.

The modal's guidance for the recordings:

• One speaker, no music or echo.
• A natural narrating pace.
• Read in the language you will narrate in — the clone keeps your accent.
• Only clone a voice you have the right to use.

Cost: Cloning is a one-off charge. The confirm button shows the price, so you see it before you commit your recordings.

Skills

Settings > Skills is where you create, edit, organise into sections, and delete your skills — the standing instructions the AI follows in every chat where you switch them on. It is described in full under "Skills".

API Keys

API Keys holds the keys for external services such as Vast.ai GPU rental. For a key you can:

• Add or change it.
• Show or hide it.
• Check the balance.
• Open the service page.

GPU Config

GPU Config controls the rented GPU search. The GPU family itself is chosen in the Available Instances modal (Create instance); GPU Config lists the selectable families — 3090 / 4090 / 5090 / L40S / A100 / RTX 6000 / H100 / H200, with RTX 5090 as the default — and sets the shared Vast.ai search limits:

The search also applies floors you cannot lower: 23 GB of GPU memory, an Ampere or newer card, 100 GB of disk, and 32 GB of host RAM.

Tip: If offers are too expensive or none appear, adjust these settings. The Change settings button on the modal's empty state opens this section directly.

Captions

Global Settings > Video Config supplies the caption defaults for newly created projects only. To change captions for an existing project:

1. Open the right scene sidebar's Settings tab.
2. Find Captions and set Burn-in captions to On or Off.
3. Apply the changes before continuing generation.

When captions are enabled, you can adjust:

• Position.
• Font.
• Size.
• Text color.
• Highlight color.
• Outline color.
• Optional background color.

Caption Fonts

The font catalogue has 51 faces, grouped by the script they are made for:

• Latin & Cyrillic (30): Bangers, Montserrat, Bebas Neue, Anton, Luckiest Guy, Poppins, Archivo Black, Lilita One, Titan One, Bungee, Paytone One, Alfa Slab One, Passion One, Kanit, Sigmar One, Bowlby One, Righteous, Carter One, Londrina Solid, Russo One, Rubik Mono One, Seymour One, Play, Black Ops One, Permanent Marker, Chewy, Amatic SC, Lobster, Creepster, Press Start 2P. (Russian is covered by Montserrat, Russo One, Rubik Mono One, Seymour One, Play, Amatic SC, Lobster, Press Start 2P, and the Japanese fonts.)
• Japanese (6): Dela Gothic One, M PLUS Rounded 1c, Zen Maru Gothic, RocknRoll One, Reggae One, DotGothic16.
• Korean (5): Black Han Sans, Do Hyeon, Jua, Bagel Fat One, Gasoek One.
• Arabic (4): Tajawal, Lalezar, Almarai, Rakkas.
• Hindi (6): Mukta, Hind, Khand, Rozha One, Yatra One, Kalam.

The default is Bangers. Inter is not in the catalogue. When the chosen font cannot write a project's language, that language's own default is used instead — both in the project picker (which says so under the field) and at burn time, instead of drawing empty boxes:

The two font pickers differ:

• The Font picker in Settings > Video Config lists all 51, each option drawn in its own typeface, with a small grey hint (Japanese, Korean, Arabic, Hindi) on the fonts made for a non-Latin script. Under it: A project in a language this font cannot write uses that language's own default instead.
• The per-project picker in the right sidebar's Settings tab lists only the fonts that can write that project's language.

Previews load the faces from the internet on first use; offline, previews fall back to the app's own font and nothing else changes.

Other caption rules:

• Caption size is independent of the resolution: the same font size takes the same share of the frame at 720p and 1080p.
• Japanese speech is split into one-line cues at phrase ends rather than by word count.
• Right-to-left languages highlight the current word on the correct side.
• Sentence ends recognise CJK, Arabic, and Devanagari punctuation.

How Captions Are Made

Captions are added after a scene's video is rendered, not during it. Each scene gets a second box on the board — raw clip → captioned clip — and the captioned one is produced like this:

1. Aificient transcribes the scene's narration audio to get word-by-word timings aligned to where the voice sits in the final video. This is a small paid step, billed like other generations.
2. Before every burn, the app self-checks the stored timings. Timings that do not match the narration, or that predate a timing change, are corrected or re-transcribed automatically.
3. Your own machine burns the captions into a copy of the clip. This local render shares the same queue as the final stitch, so it never competes with video generation on the GPU.
4. Both versions of the scene are saved to your project — the plain clip and the captioned one — so you can open the project on another computer and still see and use the captioned cuts.
5. When captions are on, the final video is stitched from the captioned clips, so the captions are baked into the video you publish.

Cost: The word timings are saved too, so re-burning a scene's captions later — to try a different font or color, say — is free. You are charged again only when something changes the narration itself (a new script, new audio, or a different voice), because that makes the old timings wrong.

Turning Captions On or Off Later

Captions can be switched at any time from the selected project's right-sidebar Settings tab. The raw scene videos do not have to be regenerated.

Because turning captions off deletes the word timings, turning them back on later re-transcribes each scene (the same small per-scene charge). That also makes an off-and-on round trip a clean reset if captions ever look wrong.

In some restored or older projects, the Stitch node can instead show a Re-stitch with captions or Re-stitch without captions shortcut when the existing final video's recorded caption mode does not match the current setting.

Note: Re-rendering a scene's video replaces its plain cut, and its captioned cut is removed at the same time — the old captions were burned onto the take you just replaced. The scene's captioned box goes back to empty, ready to be generated again through Resume Generation.

3. Publishing the Final Video

Once a project has a final video, you can post it straight to social media from inside the app. Open the Publish dialog with the Publish button at the bottom of the right scene sidebar; it becomes active only after the final video exists. Publishing is offered for History projects; posters and UGC ads are downloaded and posted by hand.

Publishing runs entirely on your own machine. The app drives a private background browser session that is signed in as your account; nothing about your accounts or your post is sent to an Aificient server. TikTok is supported today — Instagram Reels and YouTube Shorts appear as tabs but show a "coming soon" panel.

Before you can post, connect at least one account in Settings > Social Accounts (see "Social Accounts"). If you have none connected, the dialog links you straight there; you can also reach it any time with the Manage link.

Choosing a Platform

The top of the dialog has a tab per platform (TikTok, Instagram, YouTube). Each platform has its own composer, because every platform posts under different rules. Pick a platform, then fill in that platform's post. Only TikTok's composer is active right now.

Step 1 — Compose

1. Under Accounts, pick the connected account(s) to post to. Only the selected platform's accounts are listed; each shows its avatar and follower count, and selected ones get a check badge.
2. Under Caption, write your caption in the single text box and put your #hashtags right inside it (hashtags are highlighted as you type). YouTube also shows a separate Title field.
3. Or click Generate — the atom button next to the section title — to have the AI draft a caption from your project. It follows the optional Caption Instructions you set in Settings > Model Config.
4. Under Sound, attach a TikTok sound, or leave it unset to keep your video's original audio.
5. Check the live Preview on the right. It mirrors how the post will look in the feed — the cover, caption, account handle, and the chosen sound spinning on the cover disc.
6. Click Prepare publish to slide to the review step.

The sound picker lets you search TikTok's public sound catalog, spot Trending sounds, and preview any track with the ▶ button before choosing. Once a track is selected it collapses into a row with a volume slider, and it plays over your video in the live preview so you can hear the mix. Browsing sounds needs no login.

Step 2 — Review

The review step is a read-only recap plus the last two choices:

1. Check Posting to, Caption, and Sound — a summary of what you set in compose.
2. Choose the Thumbnail (cover), the exact frame used as the cover. Drag the film-strip timeline to scrub anywhere in the video, or use the left/right peeks to nudge frame by frame. The chosen frame shows as a "Cover" preview.
3. Choose When to publish: Publish now or Schedule. Scheduling reveals a date/time picker; the time must be at least ~20 minutes in the future.
4. Click Publish now / Schedule to start the hand-off, or Back to return to compose.

Upload Progress

As soon as you confirm, a progress screen takes over the dialog and locks the form so nothing changes mid-post. It shows the platform badge in a spinning ring, the current step, the account, and elapsed time, plus a row of step icons that fill in (and turn green) as the upload advances:

1. Preparing the video.
2. Opening TikTok Studio.
3. Uploading.
4. Caption.
5. Cover.
6. Sound.
7. Publishing or scheduling.

A Cancel button stops the background browser hand-off.

The progress is saved app-wide, not just in the dialog. You can close the Publish window and keep working — the post keeps going, and reopening Publish resumes the same live progress until it finishes.

Publish History

The History button (top-right of the dialog) lists past and prepared posts, with links to live posts and the option to cancel a queued one. It is hidden while an upload is in progress.

Note: TikTok publishing automates the real TikTok Studio composer in the background, so your video is uploaded with the caption, cover, and sound attached, and the post is grouped under the chosen sound. TikTok changes its site often; if a step cannot be completed automatically, the upload stops and shows an error — try again.

4. Updates

App Updates

The desktop app checks for a newer version at startup, every 15 minutes, and immediately whenever the server says the build is too old. The check runs in the sign-in window too, signed in or not. An update is either optional or required:

A dedicated update window then goes through these states:

• Preparing update, then Downloading version <x> with progress.
• Opening installer — The installer is ready. This client will close now.
• The app closes so the installer can continue.
• If it fails: Update failed / Unable to install the update. in the update window, or Unable to start update in the modal.

On the web a banner does the same job, on the sign-in page too:

• Optional: Aificient Studio <x> is available / Refresh the page whenever you are ready to pick up the latest version. with Refresh and a dismiss ×.
• Required: An update is required to keep using Aificient Studio / Version <x> has been released. Refresh the page to continue — finish anything in progress first. with Refresh and no way to dismiss it.

Updating does not sign you out: the stored session is reissued for the new version on the desktop and on the web.

What's New Cards

New features announce themselves with a What's new card. It opens by itself shortly after sign-in, last of the startup surfaces — after the introduction tour has decided whether it runs and after the daily-credit popup has had its turn — and never on top of another dialog. If something else is open, it waits for a quiet moment.

The card shows a picture, the What's new label, a title, a short description, and a few highlights.

One button closes it: Got it, or, when the announcement points somewhere, a button that takes you there — a Settings section, a composer mode on the home screen, or a link opened in your browser. The corner ×, Esc, and a click outside close it too.

Note: Each card is shown once per account. It is marked as seen the moment it appears, so it does not come back — on this device or any other — and there is no way to reopen one you have dismissed.

5. Downloads and Sharing

You can download final videos and many generated assets. When downloading:

• Choose where to save the file.
• The app writes only to the selected location.
• Downloads may take time for large videos.

The fullscreen viewer also includes a share action when your system supports it.

6. Where To Find UI Options

Use this section to answer questions such as "Where is X?", "How do I open X?", "Where can I change X?" or "I cannot find X": give the shortest path from a stable area of the app, such as left sidebar > Settings > GPU Config. If an option only appears after selecting a project, opening a modal, generating an asset, or entering render settings, say that first.

Main Navigation Locations

Creation Home Screen Locations

Render And Generation Locations

Project Review And Editing Locations

Global Settings Locations