Support & Workflows

1. Support Assistant (Ask AI)

The Support Assistant is an in-app help chat that answers questions about using Aificient Studio. Its answers are based on this user guide.

Where: the Ask AI button in the top-right window-actions area, next to the window controls.

Opening and Closing

• Open it with Ask AI. The chat is closed by default.
• Close it with the close button in the chat header, by pressing Esc, or by clicking Ask AI again.
• Reopening during the same session keeps your conversation. Closing the app clears it.

Asking a Question

1. Type your question in the message box.
2. Press Enter to send, or Shift+Enter to add a new line.
3. Wait a few seconds while the assistant prepares an answer.
4. Read the answer, then ask follow-up questions if needed.

Limits and behaviour:

• A message can be up to 4000 characters. A counter appears as you get close to the limit.
• The send button is disabled while an answer is being generated.
• When the chat is empty, suggested questions are shown. Click one to ask it right away.

What It Can Help With

• How features work, such as creating projects, characters, scenes, and videos.
• Where to find settings and controls.
• Steps for common workflows.
• Render settings, GPU options, and asset editing.

Note: The assistant explains how to use the app. It does not change your project, start generation, or perform actions for you. To have the AI change the open project, use Edit with AI in the bottom-right corner of the canvas (see "Edit with AI").

If the Assistant Cannot Answer

• If a request fails, the chat shows an inline message with a Try again option. Your question is kept so you can resend it.
• If help chat is temporarily unavailable, wait a little while and try again.
• If your session has expired, sign in again and reopen the chat.

2. Troubleshooting

Each entry names a symptom, why it happens, and what to do. Entries run by theme: sign-in, GPUs and runtimes, cloud and credits, generation and canvas, characters and assets, voices, captions, posters, UGC ads, Edit with AI and skills, publishing, and other.

I cannot sign in

Why: A connection problem, a sign-in interrupted or finished somewhere else, or an app version that needs updating.

Fix:

1. Try a different sign-in method. If you use email, request a new code.
2. Match the message to the table.

"No devnet access"

Why: Only the Aificient test environment (devnet), used by test builds, shows this: an account there stays locked until it is approved. The released apps never show it.

Fix: The No devnet access window, with your email as subtitle, reads This account doesn't have access to the Aificient devnet. To request access, write to [email protected] from the address you signed in with. It cannot be dismissed; click Copy email to copy the address, or Sign out.

No GPU is available

Why: No runtime is ready, or no rented offer matches your filters.

Fix:

1. Check the local runtime is installed and ready, with at least 24 GB of VRAM on the GPU and 24 GB of RAM on the machine. The sidebar says Needs 24 GB VRAM, the detail view insufficient VRAM, or the RAM message appears in the Error block when you start it.
2. For rented GPUs, check the Vast.ai key is configured, an instance is running and ready, and the search filters are not too strict.
3. Or pick Aificient Cloud on the initial render-settings screen, or under Resume Generation for an existing project whose images and audio are ready. Both need an active plan, enough credits, and eligible scene inputs.

Local runtime will not install

Why: A system, hardware, or disk requirement is not met.

Fix:

1. Check your system is supported (Windows desktop app).
2. Check the minimums in the table. A smaller card or less RAM means Aificient Cloud or a rented GPU.
3. Check free disk space: 40 GB for the install, plus the model download of around 76 GB.
4. Open runtime details and review the visible error/logs (Copy logs copies them).

A render is refused because the GPU is too small

Why: The selected runtime's card is plainly under 24 GB: <GPU> has less than 23 GB of VRAM, which MiniMax-H3 cannot render on even with every block streamed. Use Cloud Render, or pick a 24 GB+ GPU (RTX 3090 / 4090 or better).

Fix: Pick Aificient Cloud or a 24 GB+ GPU.

The app warns that 1080p will be slow

Why: You queued 1080p onto a GPU under 40 GB, which has to stream the model. Clips past about 6 seconds may not fit at 1080p on such a card.

Fix: choose in the 1080p will be slow on <card> dialog:

• Switch to 720p — saves the project at 720p (much cheaper, very similar quality) and queues it.
• Keep 1080p — goes ahead.
• Don't ask again — silences it on this computer.

Rented GPU offers do not appear

Why: The Vast.ai key or balance is missing, or the filters exclude every offer.

Fix:

1. Check the Vast.ai key and the account balance.
2. Increase max price, lower minimum reliability, or reduce the disk requirement (the search never goes below 100 GB).
3. Pick a different GPU family (the chevron under the featured cards shows four more).
4. Change settings on the empty-offers state opens Settings > GPU Config without closing the modal.

Aificient Cloud is unavailable

Why: Cloud needs an active subscription with credits, eligible scenes, and storage under its allowance.

Fix:

1. Initial project: check Aificient Cloud appears in the render-settings Runtime list. Existing project: confirm it has missing scene videos and its scene images and audio are ready, then check Resume Generation.
2. Check the subscription is active with enough credits for the displayed estimate. With fewer, only the first scenes the balance covers are sent; the rest wait for more credits.
3. Check the submission has no more than 32 missing scenes, and that each scene's audio plus voice delays is no longer than 13 seconds.
4. Check Settings > Storage is not over its allowance.

Note: If a queued or running cloud render fails, open Generation assets > Aificient Cloud for its status. Failed and cancelled cloud jobs refund their reserved credits.

Storage limit reached

Why: The account is over its plan's storage allowance. Generation actions that would add assets can stay blocked until it is under the allowance or the plan is upgraded.

Fix: Open Settings > Storage to see usage, the plan limit, and available space. Remove unneeded project data, or view the available plans for more storage.

I cannot claim my daily credits

Why: Daily credits need an active plan or free trial and can be claimed once per calendar day; an account with no subscription has nothing to claim.

Fix:

• Wait for the reset at 00:00 Europe/Madrid (CET / CEST), not 24 hours after the last claim. Once claimed, the card and popup disappear and the out-of-credits card shows Free credits in N hours instead.
• If you claimed on another device, the app re-syncs silently.
• Any other failure shows Unable to claim your daily credits. Check your connection and try again from the sidebar card.

The free trial checkout refused my card

Why: Prepaid and disposable cards cannot start a trial, and a card that already started a free trial on another account cannot start a second one.

Fix: Click Try another card on the result card. The trial stays available.

My plan is billed through the App Store

Why: A plan bought in the iPhone app can only be changed or cancelled in your Apple subscriptions, so web checkout and Stripe actions are refused with Your plan is billed through the App Store. Manage it in your Apple Account's subscriptions.

Fix: Click Manage in App Store (Settings > Usage or the plans modal).

Generation is incomplete

Why: One or more scenes failed.

Fix:

1. Check the generation error summary box in the top-right of the canvas for the specific failures and their reasons.
2. Open the Schema tab and find the scene with an error.
3. Open the Assets tab to see what is missing, and the runtime detail view to check logs.
4. Use Resume Generation.

Editing is disabled

Why: Editing is blocked while generation is active.

Fix: Stop generation or wait for it to finish, then edit.

The sidebar says the project is generating on another device

Why: Another computer, the web, or the iPhone app is generating this project (Generating on <place> N%). A run started here meanwhile is refused with This project is being generated on <place>.

Fix: Wait for it, or stop it from that device.

The final video disappeared after I changed a transition

Why: Expected: a transition lives in the final cut, so changing one deletes only the final video. Every clip is kept.

Fix: Click Stitch Final Video at the bottom of the sidebar to rebuild it. No scene renders again.

I added a character but cannot generate its image

Why: A new character is in no scene until you cast them, and nothing regenerates when you add them.

Fix:

1. To render the reference image for preview, use Regenerate images + audio in the confirmation right after adding, Edit > Regenerate image on the card, or right sidebar > Resume Generation > Images + Audio only.
2. The button stays active for an uncast character with no image even when the project's final video already exists. The card shows the progress, and the image lands when the run finishes.
3. To put the character in the video, use the Cast strip on the scenes they belong in.

I cannot edit, delete, or dress a character

Why: A card with a global badge is a library character and cannot be changed from inside a project (Global characters cannot be edited from the jamboard). Every card's actions are also hidden while the project is generating.

Fix:

1. Click Project variant on its card for an editable copy for this project, or edit the original in Character list.
2. If the project is generating, wait for the run to finish or stop it.

A deleted character or asset still shows up in a new scene image

Why: The scene's text still describes it, so the image model draws it from the words.

Fix:

1. Use Rewrite N scenes in the Rewrite without <name> offer that appears right after the delete.
2. If you chose Leave as it is, rewrite the scene's text yourself, or open its cast or asset picker, untick what should leave, and Save and rewrite.

"This scene is still being rewritten from your earlier save"

Why: An earlier rewrite of that scene is still running (usually after a network error). Nothing is charged twice.

Fix: Close the card or wait. It switches to the result as soon as the new text arrives.

I cannot attach more assets to a story

Why: Asset limits are fixed by the service and cannot be changed.

Fix:

1. In a story request, remove one from the Assets panel to add another.
2. In a project (at most 5 assets per scene and 3 per character), unselect one in the picker to make room.

The uploaded asset got a different name, or the upload says the AI couldn't read it

Why: Uploads are named and described by the service from the picture (a small charged call); the file name is never used.

Fix:

1. Edit the name or description in the asset's Details and Save.
2. The AI couldn't make out that image. Try a clearer picture, or try again in a moment. means the picture could not be described and nothing was stored. Try a clearer image.

"Cannot update assets while a task is in progress"

Why: Asset, outfit, and cast changes are blocked while a generation task runs.

Fix: Wait for it to finish, or stop the generation, then try again.

An asset tile is dashed / a chip says "pending generation"

Why: That asset has no image yet, usually one the script planned. It renders automatically at the start of the next generation run (before characters and scenes); a failed asset never fails the run, the scene generates without that reference.

Fix: To render it now, open its tile and click Generate image, or use Resume Generation > Images + Audio only.

The narrator does not use the voice I picked

Why: The service switched the voice while creating the project, typically because the voice does not narrate in the script's language. The Narrator card on the canvas shows a notice saying so, and the narrator is on the automatic voice for that language.

Fix:

1. Open the Narrator card and click Change voice.
2. Pick a voice native to that language (the picker opens on them), or one of the multilingual Global voices.
3. The Automatic voice accent rows in Settings > Audio Config decide which Spanish or English accent an automatic narrator speaks with. A voice name with a small Auto tag is the one the service picked automatically; play its preview to hear it.

The voice picker says No <Language> voices on the shelf yet

Why: The Narrator card's picker lists the library voices native to the script's language, and none is on your shelf.

Fix: Click Browse <language> voices to open the Voice library on that language and pick (or save) one. Picking a voice there also saves it under My voices.

I cannot create another voice

Why: Your plan caps the voices you keep (Standard 5, Pro 10, Max 20). Settings > Voices shows N/M beside the heading, where M is that cap.

Fix:

1. At the cap, the create tile becomes Upgrade to keep more voices and opens the plans modal. On the top plan, delete a voice to make room for another.
2. A tile disabled with Voice creation is not available on this server means cloning is not configured there.

My cloned voice says "Rebuilding" or "Rebuilt on next use"

Why: The voice service dropped the model; nothing of yours is lost.

Fix: Nothing to do. It is rebuilt automatically the next time the voice narrates, and the voice stays selectable meanwhile.

I cannot save another library voice

Why: The shelf of saved library voices has a cap set by the service (You already keep N voices. Remove one to save another.).

Fix: Remove one from Settings > Voices > Saved library voices, or with Remove from saved voices in the Voice library. Scripts that already use a removed voice keep it.

Captions use a different font than the one I chose

Why: The chosen font has no characters for the project's language, so the language's default is burned instead. The project's Font field says <Font> has no <Language> characters, so captions use <Font>. underneath.

Fix: Pick a font from that language's group. The project picker lists only fonts that can write it.

I cannot add another poster reference image

Why: Reference images and pinned characters share one budget of 5, and characters have their own limit of 3. The Reference assets pill shows the combined count, and its tooltip explains what is blocking you.

Fix: Unpin a character to free a slot, or remove an image.

A poster plan will not create

Why: Something the plan relies on changed, or it took too long to generate.

Fix: match the message:

The poster render button is disabled

Why: A render needs saved settings, a runtime, and work left to do.

Fix: check, in order:

1. The Settings tab has no unsaved changes. A render always uses the saved video config, so the footer warns Unsaved video settings and blocks the submit until you Apply changes.
2. A runtime is selected in the Resume Generation menu (one GPU, or Aificient Cloud when it is available).
3. The animation is not already rendered; the button is disabled when there is nothing left to do.

My poster has no captions or Publish button

Why: Posters have neither. There is no narration track to transcribe, so there is no caption step, and the Publish dialog is available for History projects only.

Fix: Nothing to do; this is by design.

A UGC ad will not render

Why: Something is missing or unsaved; the sidebar's Resume Generation tooltip or the clip's Render tooltip names it.

Fix: match the tooltip:

I cannot add another photo, person, or place to a UGC ad

Why: The ad is at one of its limits:

• 6 photos of the product, its assets, and its places together.
• 2 people and 6 places.
• 10 reference images in all; a scene sends at most 5.

Fix: Delete a photo or a place you no longer need (Delete asset / Delete location in its modal), or take an image out of a scene. Photos with people in them are not used as assets; add people with Add character.

A UGC clip says "The read slipped"

Why: When a clip lands, the service transcribes it and compares it with the line. The strip under the clip lists the words that differ (montaña → montañaña, +la for an extra word, −siempre for a missing one). It is a warning only; nothing re-renders and nothing is charged.

Fix:

1. Hover the strip to read every mismatch, then play the clip; the transcriber can raise a false alarm.
2. If the slip matters, click Regenerate on the clip for another take (a new render is charged).
3. If only the start or the end is wrong, the scissors on the clip node can trim it out of the final video for free.

My UGC ad's texts or captions are not in the final video

Why: Texts and captions are burned into the scenes' finished cuts, which are made when the ad is finished. Emoji are always left out.

Fix:

1. After changing them, use Stitch Final Video (or Finish the video).
2. If a burn fails, that scene falls back to its plain clip and the report says Scene N: <error> — its texts are not in the final video.

Edit with AI did not do what I asked

Why: A step failed, or the assistant refused or is waiting for your answer.

Fix:

1. Read the timeline under its answer: each step it ran is listed, and a failed one is red with its reason.
2. It refuses media changes while a generation runs on this project (here or on another device): wait for it, or ask it to stop the generation.
3. It always asks before rendering video or anything destructive, and waits for your answer in the chat.

My skill is not being followed

Why: It is switched off, or the project's assistant uses its own selection.

Fix:

1. Check the Skills pill, which shows how many are on (a sent message shows a puzzle glyph with the count). At most 10 can be on at once.
2. In a project's Edit with AI, the assistant has its own selection, which starts from the skills the project was planned with.

Publishing to social accounts

For the full walkthrough of the Publish dialog (compose, review, scheduling, and live upload progress), see "Publishing the Final Video".

Why: A missing final video, an account that is not connected or has expired, or a TikTok page change mid-upload.

Fix: match your case:

Note: Publishing runs entirely on your own machine through a private background browser session. Nothing about your accounts or posts is sent to an Aificient server.

Rename project fails

Why: The runtime or cloud API that owns the project is unreachable.

Fix: Rename from the project list context menu. If it fails, check that the owning runtime or cloud API is reachable, then try again.

The what's new card did not come back

Why: Each card is shown once per account and is marked as seen the moment it appears.

Fix: Nothing to do. It cannot be reopened, on this device or any other.

The introduction tour did not appear

Why: It starts on its own only for an account less than 14 days old, on the home screen, once billing has loaded, and only once per device or browser.

Fix: Replay it any time from Settings > General > Introduction tour > Replay.

3. Recommended Workflows

Step-by-step paths through common jobs; each feature's own section has the details.

Fast Preview Workflow

Use this when you want to inspect assets before video rendering.

1. Create a concept.
2. Choose Generate images and audio only.
3. Review images and narration.
4. Rewrite weak scenes or characters.
5. Resume generation later with a ready GPU.

Full Video Workflow

Use this when you want a final video immediately.

1. Create or select a concept.
2. Choose aspect ratio and resolution.
3. Select one or more ready GPUs (pick several to split the render and finish faster), or choose Aificient Cloud by itself when available.
4. Choose Lite for speed and lower VRAM, or Pro for best quality.
5. Start generation.
6. Monitor the canvas.
7. Download the final video from Output.

Scene Fix Workflow

Use this when only one scene needs changes.

1. Select the scene.
2. Click the text, image, audio, or video node you want to change.
3. Rewrite or delete the asset.
4. Resume generation.
5. Review the updated scene and final output.

Character Reuse Workflow

Use this when you want consistent characters across projects.

1. Create a character in Character mode (it lands in the library), or save a project character from its card with Library.
2. Pin it when creating a new concept, or add it to an existing story from Add character > Library.
3. Generate the project.
4. Review scenes where the character appears.
5. If the character needs a project-only change (a different outfit, a scar, an older look), click Project variant on its card and edit the copy.

Add a Character Mid-Project Workflow

Use this when a story needs someone who was not in the original script.

1. Click Add character under the cast column. Describe them (Manual), let the AI write them from a prompt or reference photos (AI), or pick one from Library.
2. Click Regenerate images + audio in the confirmation (or Resume Generation > Images + Audio only) to render the reference image and check the look. Nothing else regenerates yet.
3. Adjust with Edit if needed. Keep the current look as reference keeps the face while you change details.
4. Open the Cast strip on each scene the character belongs in and Save cast. Those scenes regenerate their image and video on the next run.
5. Resume Generation to rebuild the affected scenes and the final video.

Brand or Product Placement Workflow

Use this when a real product, logo, or location must appear consistently.

1. In the Story composer, attach the packshots and location photos with the Assets pill (up to 8), then describe the video.
2. Read the plan card's Assets block and the asset chips on each scene row. Ask for corrections in chat if the product should appear elsewhere.
3. Generate. Uploaded assets are never regenerated: what you uploaded is what the image model is shown.
4. On the canvas, fine-tune: Edit assets on a scene to add or remove the product, or open the asset's tile and use Place with AI (show it wherever the chemist is drinking).
5. Missing something? Upload an image or Create from a prompt on the Assets card adds new assets at any time.
6. Resume Generation to rebuild only the scenes you changed.

Consistent Wardrobe Workflow

Use this when a character must wear the same outfit in every scene.

1. Add the outfit to the asset base (upload a photo of the garment, or Create from a prompt with kind Wardrobe), or attach a reference photo when adding the character.
2. On the character card, Add outfit and select the asset (up to 3 per character). Saving regenerates the character's image and the scenes they appear in.
3. Resume Generation > Images + Audio only to review the new look before spending render time.
4. Resume Generation with a runtime to re-render the affected clips and the final video.

Swap the Narrator Voice Workflow

Use this when the script is right but the narrator is not.

1. On the canvas, open the Narrator card and click Change voice.
2. Pick a voice native to the script's language, a Global voice, one of My voices, or Automatic. Or Browse thousands more voices and pick one from the Voice library (it is saved to your shelf at the same time).
3. Read the Change narrator confirmation: it tells you how many scenes are re-recorded and the approximate credits. Images and rendered clips are kept.
4. Click Change and re-record. The narration is re-recorded straight away.
5. Resume Generation to re-render only the clips whose new narration no longer fits, and to stitch the final video again.

Animate an Existing Poster

Use this when you already have finished artwork.

1. Open Poster mode and select I have a poster.
2. Attach the poster image and describe how it should move.
3. Pick a clip length, then send the prompt.
4. Refine the plan in chat until the motion reads right.
5. Create project, choose Lite/Pro, resolution, and one runtime, then Start generation.
6. Download the animation from the final-video node.

Design and Animate a Poster From Zero

Use this when there is no artwork yet.

1. Open Poster mode and select From zero.
2. Describe the poster, attach up to 5 references (each pinned character counts as one), and pick the aspect ratio and duration.
3. Review the plan's three prompts and refine them in chat.
4. Create project with Images only if you want to inspect the poster before spending render time.
5. Check the designed poster and first frame on the canvas. Rewrite a prompt and regenerate the images if needed.
6. Use Resume Generation > pick a runtime > Add to render queue to render the animation.

Product Ad From a Shop Page

Use this when you sell something online and want a creator-style ad for it.

1. Open UGC mode. In the Assets pill, paste the product's page into Product URL and Import it. Open the product card to check the photos, the name, the price, and the key points.
2. Optionally add photos of the places to film in and mark them Location, pin a creator with Characters (the first one leads), and pick a Voice and an Angle.
3. Describe who it is for and the offer, and send. Refine the ad plan in chat ("punchier hook", "add a demo shot", "shorter lines").
4. Generate, then turn on Generate storyboard only and Create storyboards to check every shot as four sketches before paying for video.
5. On the board, fix what needs fixing: open a scene to change its line, camera, or place, play its read line, or Redraw a storyboard.
6. Resume Generation > Generate video. Render a complex scene on Pro from its clip node if Lite is not enough.
7. Add on-screen texts in the scenes' Texts tab and pick the transitions, then Stitch Final Video and download the ad from the Output node.

Crazy Hook Ad

Use this when you want an ad that stops the scroll.

1. Open UGC mode and switch on Crazy hook. Pick a Hook idea (or leave Surprise me) and a length.
2. Attach clear photos of the product (no people) and describe what you sell and who it is for.
3. Read the Hook block of the plan (the event, the reveal, and the twist) and refine it in chat ("a bigger event", "a faster reveal").
4. Generate with Max if the chaos needs more believable physics, or render the key scene alone on Max.

Edit a Project by Asking

Use this when you know what you want changed but not where the button is.

1. Open the project and press Ctrl+K (⌘K on a Mac), or click Edit with AI in the bottom-right corner.
2. Ask for the change in plain words, several at once if you like ("regenerate the image of scene 2, make scene 3's narration more dramatic, and turn captions on").
3. Read the timeline under the answer to see what it did. Before rendering video or anything destructive it quotes the cost and waits for your yes.
4. Keep working on the board while it works; Stop cancels the current turn.

Keep Your House Style With Skills

Use this when every video should sound and be structured the same way.

1. Open Settings > Skills > New skill and describe the rule in your own words ("open with a bold question, short sentences, end with a call to action to my site"). The AI writes the skill and names it.
2. Group skills into sections (Brand, Kids, UGC…) if you keep several.
3. Switch them on from the Skills pill before you send a brief. They stay on for the next chats on this device until you switch them off.
4. In a project's Edit with AI, the skills the project was planned with are already on.

Polish the Cuts With Transitions

Use this when the scenes are right but the joins feel abrupt.

1. On the canvas, find the Transition node between the two scenes and click Change transition.
2. Hover the tiles to preview them, pick one, set its Length (and direction for Slide or Wipe), and Apply.
3. Repeat for the other cuts, or ask Edit with AI to suggest transitions.
4. Stitch Final Video: only the final cut is rebuilt; no scene renders again.

4. Glossary

Terms used across this guide, in alphabetical order.

Aificient Cloud

The subscription-backed render service, available on initial render settings and from Resume Generation for eligible scene videos. It uses credits and has its own queue, separate from local and rented GPU queues.

Asset

A generated file such as an image, audio clip, scene video, or final video. Not to be confused with a project asset (see "Project Asset / Asset Base").

Automatic Voice

The narrator setting that lets the service choose the library voice for the script's language and accent, honouring the Automatic voice accent rows in Settings > Audio Config. Shown as Automatic on the Narrator card and in the pickers.

Cast

The characters that appear in a scene, shown as chips on the scene card and edited with Edit cast. Changing a scene's cast regenerates its image and video; narration is kept.

Cloned Voice

A narrator voice created from your own recordings in Settings > Voices; how many you can keep depends on your plan. It appears under Cloned in the My voices tab of every voice picker, and any scene whose narrator names it is narrated with it.

Crazy Hook

A UGC ad that opens on something impossible, an absurd event filmed like real footage, that reveals what you sell. A switch in UGC mode, with its own Hook idea and lengths of 10–30 seconds.

Daily Credits

A free top-up that accounts on a plan or trial can claim once per calendar day (reset at 00:00 Europe/Madrid) from the sidebar card, the daily popup, or the out-of-credits card. It always takes a click.

Edit with AI

The in-project assistant in the bottom-right corner of the canvas (Ctrl+K / ⌘K). It changes the open project when you ask, runs the same actions as the buttons, asks before anything costly, and can spend credits; not to be confused with the Support Assistant.

Final Output

The completed video shown in the Output node.

Finished Cut

A UGC scene's clip with its captions and on-screen texts burned in. The final ad is stitched from the finished cuts (or the plain clips where none is needed).

First Frame

The opening image of a poster animation, derived from the finished poster. The animation runs from the first frame back to the poster.

Free Trial

A time-limited trial of one plan: save a card, receive the trial's credits at once, pay nothing until the trial ends. Its buttons carry a gift-box glyph.

Generation Run

A record, kept by the service, of one project being generated on one device (this app, the web, or the iPhone app). Other devices show it as a Generating on <place> N% chip and wait for it; only the device running it can stop it.

Global (Library) Character

A character that lives in your library and is referenced by a project rather than owned by it. It shows a global badge on the canvas and cannot be edited, deleted, or dressed from inside a project.

Guidance (CFG)

Classifier-Free Guidance: a setting that controls how strictly the AI follows your prompt, where higher values mean stricter adherence and lower values allow more creative freedom. Video guidance (Video CFG) applies to visual generation, and Audio guidance (Audio CFG) applies to sound generation.

Headroom

The third bar on a GPU rental tier card: how much of the model stays on the card instead of streaming over PCIe each step. Cards under about 40 GB stream and are slower, especially at 1080p.

History Project

The multi-scene kind of project: a concept becomes a script, scenes, images, narration, clips, and a stitched final video. It is what the Story chat produces (including when it brainstorms first).

Introduction Tour

The seven-step guided walk through the home screen that starts for new accounts and can be replayed from Settings > General.

iPhone App

The native iPhone version of Aificient: the same account, projects, characters, voices, skills, credits, and plan, with Story, Poster, UGC, and Character modes, Edit with AI, per-cut transitions, cloud rendering only, Live Activities on the lock screen, and plans bought through the App Store. GPUs and publishing stay on the desktop and the web.

Lead

The first creator pinned to a UGC ad. The lead speaks on camera, and the ad's voice is shown on the lead's card and changed from its Edit modal.

Lead-in (Before the line)

Up to 3 seconds of action a UGC scene plays before its spoken line starts, set in the scene editor. It is part of the prompt and the duration, so changing it re-renders the clip.

Live Activity

The iPhone lock-screen and Dynamic Island card that follows a project's generation step by step and opens the project when tapped.

Narration

The spoken audio generated from the scene script.

Narrator Card

The canvas card between the cast and the scenes that shows the project's narrator voice and offers Change voice. Changing the voice re-records the narration; scene images and rendered clips are kept.

Needs Review

The red state of a UGC scene whose prompt may still count on an image or place that was deleted, or that names an image it no longer sends. No video renders until it is fixed or marked Looks right.

Outfit (Wardrobe)

The project assets a character wears, shown as chips on the character card and capped at 3 per character by the service. They are attached to the image model when the character's reference image is rendered, so the look follows the character into every scene.

Poster Plan

The output of a poster chat: a title, up to three prompts (poster, first frame, video), the attached references, and the clip's duration and aspect ratio. Creating a project from it is what starts generation.

Poster Project

A project that animates one poster into a short clip (5–15 seconds) with model-generated sound. It has no scenes, captions, or stitch step; the rendered animation is the final video.

Project Asset / Asset Base

A reusable reference image in a story project (a product, a wardrobe item, a location, a prop) that scenes and characters reference by id; the asset base is the Assets card on the canvas that holds them. Assets come from the composer's Assets pill, the script planner (AI planned), Upload an image, or Create from a prompt; uploaded assets are ground truth and are never regenerated, generated ones can be re-rolled, and a pending asset has no image yet and renders at the start of the next run.

Project Variant

An editable project-only copy of a library character, created with the Project variant button. The library entry and every other project using it stay untouched.

Raw Clip

A generated scene video before final stitching.

Reference Budget

The five slots a poster's reference images and pinned characters share. Each character counts as one slot, and characters have their own limit of three.

Render Queue

The per-GPU line of pending work: each scene video is a job that waits its turn on the GPU it was assigned to, so multiple projects can share a GPU and a single project's videos can be split across several GPUs. The final stitch is not a GPU job; it runs locally in the app.

Runtime

A local or rented GPU environment used for video rendering. It runs the MiniMax-H3 engine, the same one Aificient Cloud uses, and needs a 24 GB GPU (plus 24 GB of system RAM for the local runtime).

Saved Voice

A voice from the Voice library that you bookmarked (Save voice) or picked for a field. Saved voices appear under Saved in the My voices tab of every picker, speak every language, and are listed in Settings > Voices > Saved library voices, a shelf with a cap set by the service.

Scene Image

The still image used as the visual basis for a scene.

SFX

The sound effects and ambience the video model composes into each clip's own soundtrack. There are no separate SFX tracks or cues to edit; the SFX volume setting controls how loud this soundtrack plays under the narration.

Skill

A standing instruction the AI follows in every chat where it is switched on: your tone, structure, or rules. Written by the AI from your description in Settings > Skills, switched on from the Skills pill (up to 10 at once), and kept with your account.

Read Check

The comparison the service runs between a delivered UGC clip and its line. A word that differs shows as The read slipped: … under the clip; it is a warning only, and Regenerate gets another take.

Spoken Line (Read Take)

The voice's reading of one UGC scene's line, made before the clip. The clip lip-syncs to it word for word, the scene lasts as long as the line, and you can play or re-read it on the scene card.

Stitch

The process of combining scene clips into the final video: each cloud scene's narration is mixed over its clip's soundtrack, each narrated scene is trimmed to end just after its voice, and the scenes are joined with each cut's transition (or the project's crossfade). It runs locally in the desktop app once every scene clip is ready; only one stitch or caption burn runs at a time, and the others wait in a local queue.

Storyboard

Four quick sketches of one UGC scene's shot (opening frame, two key moments, final frame), drawn for a small charge so you can check the shot before paying for its video.

Support Assistant

The in-app help chat, opened with the Ask AI button, that answers questions about using the app based on this guide.

Transition

How a scene comes in after the one before in the final video: a hard cut, a crossfade, a dip to black, a wipe, a zoom, and nine more. Set per cut on the canvas's Transition nodes; changing one rebuilds only the final video.

Trim (UGC)

A cut of up to 10 seconds from the start or the end of a UGC scene's clip, applied in the final video only. The clip stays whole, nothing renders, and Stitch Final Video rebuilds the cut; at least 1 second of every scene is kept.

UGC Ad

A short creator-style ad for a product, a course, or an event: one to ten scenes with beats, a creator on camera (or a voice off screen), rendered on Aificient Cloud from your photos and the voice's reading of each line, then captioned, given on-screen texts, and stitched. Made in UGC mode; listed under UGC ads.

Voice Library

The explorer of thousands of narration voices, opened with Browse thousands more voices from any voice picker or Explore in Settings > Voices. It filters by language, accent, gender, use case, and sort order, previews each voice, and lets you pick or save it.

Welcome Credits (Gift Trial)

The small credit gift a new account starts with, enough for one or two Lite 720p cloud clips. Renders made on gift credits carry a small Aificient watermark, and new projects default to Lite.

What's New Card

The once-per-account announcement shown at startup when a feature ships. See "Updates".

Write-in

The option in a scene's cast or asset picker (Write them into the scene, Save and write in) that has the AI rewrite the scene's description and character reference so a newly added character or asset is actually in the shot, not only attached as a reference. It works the other way too (Take them out of the scene / Save and rewrite writes a removed one out of the text, and Rewrite without <name> does the same for every scene after a delete); it is a charged scene rewrite, and no scene is added.