BOS Video AI FAQ

Use your business's BOS account (username and password). You can turn on Remember username so you don't have to type it again next time.

Each account can be signed in at one place only: signing in on this computer signs your BOS account out on any other device.

• If your account belongs to one business: the app opens that business directly, without asking.
• If your account belongs to 2 or more businesses: each time you sign in, the app shows the Choose your business page. Pick a business, then click Open this business. The business you used last time is at the top of the list and already selected.

To switch to another business, click your name at the bottom of the left sidebar, then click Switch business (only shown when your account belongs to 2 or more businesses). Each business has its own video footer details, so switching business also switches the video footer.

After you sign in, the setup assistant walks you through 3 steps:

1. Connect AI: paste your OpenRouter key and click Check key. You need a valid key and an account with credit left to continue.
2. Business details: your business name, phone number, website and industry, pre-filled from your BOS account. Check them, fill in any empty fields, then click Continue. If the name, phone number and website are all filled in, the assistant skips this step. This information appears in the footer of every video.
3. First video: pick a sample idea, then review and edit it before creating.

You can skip with Later. To open it again, go to System settings and click Open setup assistant.

From your BOS account. The business owner enters the name, phone number and website once in the business info on BOS; every employee who signs in to the app gets them automatically, with no retyping.

• For any field left empty on BOS, fill it in under System settings → Video footer (or in step 2 of the setup assistant).
• Fields you have typed or edited keep your values, even after you sign in again.
• The From your BOS account line under the footer shows which fields come from BOS.

After the business owner edits the info on BOS, go to System settings → Video footer and click Update again from BOS. The app fetches the name, phone number and website from BOS again and replaces the fields, including fields you edited. Fields that are empty on BOS keep what you typed.

1. Create an account at openrouter.ai (signing in with Google works too).
2. Add credit under Credits; a few USD is enough for many videos.
3. Go to Keys, click Create Key and copy the key, which looks like sk-or-....
4. Set a Credit limit on the key (about 2 USD) so you don't overspend.

The System settings page has three tabs:

• OpenRouter · Cheaper: one key for scripts, voices, images and low-cost AI video. Enough to get started.
• RunningHub · Higher-quality video: motion clips from photos, virtual presenters, advanced tools; runs in the cloud, no GPU needed.
• ComfyUI · For computers with a GPU: runs workflows on your own computer (or your own ComfyUI server), with no cloud fees.

The OpenRouter and RunningHub tabs have a Check key button; the ComfyUI tab has a Check connection button. Remember to click Save settings after making changes.

The Create video page has 4 steps:

1. Kind: choose how to make the video and the aspect ratio (Vertical 9:16, Square 1:1, Landscape 16:9).
2. Content: what the video is about, and a length of about 15, 30 or 60 seconds.
3. Voice & template: voice, visuals, frame template, subtitles and background music.
4. Review: check your choices, see the estimated cost (USD with a reference VND amount), the cost cap and the time, choose Script review (Auto or Manual), then click Add to queue.

The video is created on the Queue page; the Create video page clears right away (keeping the industry, video type, voice and template) so you can prepare the next video.

• The app creates videos one after another in the order you added them. By default it makes 1 video at a time; change this under System settings → Videos made at the same time (1–3). Only increase it on a powerful computer (16 GB RAM or more).
• Batch creation: step 4 shows the estimated total and the maximum total (number of videos × the cap per video); you must tick the confirmation box before you can add them to the queue. Each topic is a separate item in the queue, so you can track progress, stop or retry each video; each video has its own cap.
• The Queue page shows the video being created (progress, current step, time, cost), waiting videos and finished videos. Each card has a status label: Waiting, Creating, To review, Done, Failed, Stopped or Interrupted. Click Stop to stop after the current step (a paid step already running still finishes and is still charged), Remove from queue for a video that hasn't started, Open to review for a video awaiting review, and Retry after an error (or Continue when some AI clips are already done). Finished videos have Watch video, Open in History, Open video folder and, for product review videos, Try another idea. Remove from list removes the card from the Queue page.
• You can close the tab or open another page; when a video is done, the app notifies you once on the page you have open. Videos only start while you are signed in. If you close the app while a video is being created, that video becomes Interrupted (click Recreate); waiting videos continue when you sign in again (the card says "Waiting for you to sign in to continue").
• The queue does not fill up your disk: when a video is done, the copies of its input photos/videos are deleted right away (the finished video stays in History); finished items are removed automatically after 7 days, and only the 100 newest items are kept. Remove from queue and Remove from list also delete the input copies, without touching the videos in History.
• If you open the app in two windows at once, only one window runs the queue; the other shows "The queue is being run by another BOS Video AI window". Stopping, resuming and the number of concurrent videos can be changed in either window and apply to both.
• If your OpenRouter or RunningHub balance is lower than the estimated cost of the next video, the queue pauses instead of failing. Add credit, then click Resume queue.
• A video whose estimate exceeds its cost cap does not run (it shows an error, and nothing is charged); the queue moves on to the next video.

In step 4, choose Script review: Automatic (default) goes straight through to the video; Manual stops once the script is ready (title, narration, visual descriptions), before anything is spent on voice, images or video. Change the default mode under System settings → Script review. Presenter videos are always created automatically.

• Manual videos have the status To review in Queue and History; the queue moves on to the next video while it waits for you. Click Open to review to open the Script review page.
• On this page you can edit the title, narration and visual descriptions (for videos from your photos/videos: choose a photo or video segment for each scene, or Fill with AI image with a visual description), add, delete and reorder scenes, and see what has been spent on the script, the estimate for the rest and the video's cap. If the estimate exceeds the cap, you can't approve yet; remove some scenes or shorten the narration.
• Approve and produce moves the video to the front of the waiting videos (created right after the video currently running) with exactly the script you edited; the script is not rewritten (no extra charge), and the amount already paid for the script counts toward the cap. Cancel draft drops the video (the amount paid for the script is not refunded).
• Batch creation in Manual mode: each topic becomes its own draft awaiting review.
• The card in Queue shows how long a draft has been waiting (for example "awaiting review for 3 days"); drafts are never cancelled automatically. Drafts stay intact when you close and reopen the app. If an approved video fails or is interrupted while being created, click Continue (in Queue or History): videos from an idea, a link or a batch reuse the finished scenes at no extra charge; videos from your photos/videos show Redo production with an estimate, because the voice and AI clips are made and charged again. The resumed run gets a full new cap; what was spent before is shown as "spent last time". Remove from queue on an approved video that hasn't started returns it to To review (the paid script is not lost).

• From an idea: you describe an idea, and AI writes the narration and draws the visuals. You can create in batches: each line is a topic that becomes one video (up to 100).
• From your images and videos: use existing product photos and videos. This option currently only produces vertical 9:16 videos and uses the free Edge voice.
• Virtual presenter: a character speaks an introduction for your product. Requires RunningHub, DashScope or your own ComfyUI.
• From a link: paste a product page address and click Read page so AI summarizes it into content (if the page blocks apps, the app automatically asks AI to read it).

In step 1, choose your industry: Real estate, Travel / Hotels, Food & beverage (F&B), Education / Courses, Spa / Beauty, Affiliate marketing, or Multi-industry (the standard approach).

Each agent comes with script templates, a tone of voice and instructions specific to the industry (including advertising restrictions). Click Use this template and the app fills in the video type, aspect ratio, length, voice, frame template and music; you can still change them in later steps.

In step 2, the Quick fill section lets you fill in a few fields (fields marked * are required), then click Fill the content so the app combines them into the content sent to AI.

Some agents have extra industry-specific features:

• Read product link (Affiliate marketing): paste the product page link into the Product link field in step 2, then click Read product link (if the page blocks direct reading, the app automatically asks AI to read it, or you can click Read with AI); the app reads the name, price and description into the field below (editable). If the page can't be read, type or paste the information into that field yourself.
• Industry-aware photo/video analysis: when AI looks at your photos/videos (using a VLM model through an API), it pays extra attention to what matters in your industry, for example the number of rooms and the view for real estate.
• Illustrations only: for Real estate, Food & beverage, Spa, Travel / Hotels and Affiliate marketing, AI images/videos are only generic illustrations (people, settings, atmosphere, everyday objects). They never realistically depict the actual property, dish, room or product being advertised, and contain no brand names, logos or text on packaging. For videos from your photos/videos: AI only fills scenes that lack real footage and does not fill scenes about the property/dish/product itself. For videos from an idea or a link: every scene is an AI illustration following the rule above — if you want customers to see the real thing, make the video from your photos/videos.
• Agent-specific AI models: an agent can set its own models for writing scripts, viewing images, drawing images and making videos; step 4 shows an Agent AI models line when it does. If a model is no longer available on OpenRouter, the app uses the model from System settings.

Go to System settings, open the Industry agents tab and choose the agent to edit. You can edit the opening line of AI image/video prompts, the style of AI motion clips, the instructions for AI, default values (aspect ratio, length, voice, frame template, music, subtitles, AI image or AI video, script review), the agent's tools and AI models and its script templates (add, edit, delete templates; each template has a request for AI and a table of quick-fill fields; a field with an Options column becomes a dropdown). Click Save changes; if any field is invalid, the app tells you exactly which field to fix.

• Changes apply to videos created after you save, with no need to restart the app. Videos already in the queue keep the settings they had when added.
• Your edits are saved separately on your computer, so app updates don't erase them. To go back to the original, use Restore defaults (for individual parts or the whole agent).
• The prompt line of Multi-industry is the default for every video that doesn't have its own.
• To make your own agent: open New agent, choose an agent to copy, set a name and icon, then click Create agent. Agents you create can be deleted with Delete agent.

Choose From your images and videos, then upload photos in step 2. When you use photos only, step 3 has an AI motion clips: On / Off switch (Off: free slow zoom on photos), frame templates, subtitle styles and effects. If the list includes video files, the video uses the default frame.

The app saves the video plan (idea, script, scene list) as soon as the script is written, then saves each finished part (voice, AI images, AI clips). Reopen the app: the video automatically goes back into the queue once and continues from the unfinished step. If the video fails or is stopped, click Continue in Queue or History: the existing script, voice, AI images and AI clips are reused at no extra charge; only the rendering and video assembly (free) run again.

With the Affiliate marketing agent and the Review from product link script: paste the link in step 2, then click Read product link. The app takes the page text and up to 6 product photos declared by the page (featured image, product photos) and downloads them safely (public addresses only, only JPG/PNG/WebP images large enough). Your own photos (if any) come before the page's photos; untick Use … product photos from the page if the photos on the page don't show the right product. The page text goes into the video automatically as reference information; you don't need to click Fill the content.

Upload videos in step 2 (you can add photos too). The app automatically splits each video into short segments (2–10 seconds) at scene changes; AI looks at each segment and picks the one that fits each line of narration, so scenes use different segments and cut exactly from the start of a segment. In step 3, each video has an Original audio field: Replace with voiceover (default), Mix with voiceover (original audio quietly in the background) or Keep original audio (scenes using that video play the original audio, with no voiceover). The Scenes without a matching video field fills them with an AI image (default), an AI video (more expensive) or the least-used video segment; filling with AI requires an OpenRouter key. The cost of AI viewing each segment and of filler visuals is included in the step 4 estimate and in the cost cap; if you choose to analyze videos through RunningHub or your own server, the app analyzes at most 8 segments per video and can't estimate the cost of this part yet. Videos are split right after upload; meanwhile step 3 shows "analyzing video…". A few frames from each segment are sent to an AI service to see what they show.

The address must start with http:// or https://. While reading, the app shows Reading page… with a seconds counter and a Cancel button. The Continue/Back buttons are locked meanwhile; click Cancel if you don't want to wait. If you edit the content field while waiting, the result does not overwrite it but is shown separately with a Fill in button.

When direct download doesn't work (the page blocks apps, only shows text when JavaScript runs, or has too little text), the app automatically tries, in order:

1. Open the page in the app's built-in browser (free): reads the text on the page and the product details (name, price, brand) the page declares.
2. AI looks at a screenshot of the page if the page still has too little text (requires an OpenRouter key).
3. AI opens the page itself ("The page blocks direct reading, asking AI to read it…", requires an OpenRouter key).
4. If that still fails: the app asks you to take a screenshot of the product page and drop it into the box that appears (screenshots from the web or from apps like Shopee and Lazada both work), or paste the product description into the Video content field. Pasting by hand always works, even without a key.

Each step has its own time limit (direct download 1 minute, browser 35 seconds, AI screenshot 45 seconds, AI opening the page 1.5 minutes; the seconds counter restarts when moving to the next step), so you never wait forever. The AI part usually costs about 0.001–0.004 USD each time (a few dozen dong; the estimate is shown under the Read with AI button), and each AI step costs no more than 0.05 USD; summarizing the text that was read usually costs under 0.001 USD. The actual cost is shown after reading finishes, even if you click Cancel. You can also click Read with AI right from the start.

The app does not solve captchas, does not pretend to be a user and does not sign in on your behalf. Shopee product pages require signing in, so the app asks for a screenshot right away (it doesn't spend money trying methods that are sure to fail). Links to images or PDFs are not web pages: download the file and add it to the video as source material.

• A presenter photo with a clear face, looking straight at the camera.
• A product photo (optional).
• A presenter script: write it yourself, or let AI write it from the product name (requires a product photo and the product name).

You need to configure RunningHub, DashScope or your own ComfyUI in System settings.

• Edge TTS (default, free): Northern female (Hoài My) or Northern male (Nam Minh).
• OpenRouter voice (paid): natural Vietnamese voices, about 0.015 USD for a 60-second video.
• ComfyUI: a voice workflow running on your ComfyUI.

Use Preview voice to listen before creating a video. Change how voices are generated in System settings.

1. Open the My music page from the left menu (or click Upload new music in the Background music section of step 3).
2. Choose an MP3, M4A, WAV, OGG or AAC file (up to 30 MB, at least 5 seconds long), and give it a name if you like.
3. Tick I have the right to use this track and take responsibility for its copyright, then click Add to my music.

The app evens out the volume automatically and stores the track on your computer (it stays after updates). In step 3, your tracks are at the top of the Track list; music loops or is trimmed automatically to fit the video length and follows the Volume slider. "Product review" videos also keep exactly the track you chose instead of picking music based on the idea.

Need music? Find free music, for example on Pixabay Music, or use music you have licensed. The music built into the app (the Built-in music section) is all CC0 royalty-free music.

Delete or rename tracks on the My music page. Videos already added to the queue keep their own copy, so deleting a track doesn't break waiting videos.

The Advanced tools page (Image to video and Motion transfer) only appears in the menu once RunningHub, DashScope or ComfyUI is configured.

There are 5 built-in styles: Brand, TikTok, Bold yellow, Minimal and Big UPPERCASE. The Preview section shows what the subtitles look like before you create the video.

You can choose subtitles when making a video From an idea, From a link, or From your images and videos when using photos only.

Open Customize under the Subtitles section to change: font, position (Bottom, Middle, Top), text color, background color, bottom border color, background opacity, font size, text outline and shadow, and UPPERCASE. Background opacity 0% means no background box; in that case, turn on the text outline.

It's how each subtitle line appears during the first 0.4 seconds or so of a scene: None, Fade in, Slide up or Gentle zoom. Turning on an effect makes each scene take a little longer to render.

It's the fifth option under Entrance effect: subtitles appear in short chunks (up to about 6 words, no more than 2 lines) and the word being spoken changes to a highlight color right as the voice reaches it. Choose the highlight color in the Spoken word color field; the font, size, text color, background, outline, position and UPPERCASE follow the selected subtitle style. The Preview section draws it exactly as it will appear in the video.

• Edge-TTS voice: text syncs word by word, off by less than 0.3 seconds.
• OpenRouter or ComfyUI voice: these don't report the timing of each word, so the app estimates it from word length and sentence breaks; it may be slightly off. If the chosen voice fails and the video switches to Edge-TTS, the subtitles sync to the Edge voice automatically.
• It does not wrongly split items like "TP. Hồ Chí Minh", "3,5 tỷ", "1.200.000đ" or "shopee.vn".
• Scenes without a voiceover have no karaoke subtitles (they keep the frame's static text).
• The frame's title, price and logo stay unchanged; only the static narration text is replaced by karaoke subtitles.

Subtitles are burned into the video once, after the scenes and background music are combined, so rendering takes about one extra re-encode longer. The subtitle file subtitles.ass is saved in the same folder as the video (useful when posting to platforms that accept separate subtitle files). Some agents (for example an industry with subtitles: karaoke: true in its configuration) turn this effect on automatically when you pick a script template; you can still change it in step 3.

When you make a video from photos, RunningHub turns each photo into a clip with camera motion (keeping the architecture and interior intact). You need to enter your RunningHub key in System settings. Otherwise, the app uses a slow zoom on photos (free).

Turn it on in step 3 with the AI motion clips switch above the frame templates. Clips play inside the photo area of the chosen template: full screen for Reel, and inside the card's photo area for the main card templates (Default, Modern, Elegant, Neon…); the title, subtitles, logo and decorative frame always sit on top of the clip. Templates that can't hold a clip are greyed out when clips are on. Step 4 shows the template and AI motion clips: On/Off.

Each clip can take from a few minutes to over an hour. The card in Queue shows "Waiting for clip 2/6…" and the status of each scene. A video that waits a long time switches to Waiting for video and gives its turn to the next video; when the clips are ready, it continues automatically. Closing the app midway doesn't lose clips: each scene's task ID is saved as soon as it is submitted, so when you reopen the app the video keeps waiting for those same tasks, without submitting (and paying) again. Each video waits for clips for at most 90 minutes (counting only time actually spent waiting for clips; runninghub_clip_wait_minutes in config.yaml) and gives up its turn after 10 minutes of waiting (runninghub_park_after_minutes). While it has given up its turn, the video doesn't run heavy work (placing clips into the frame only happens once it gets its turn back).

One clip per scene, with the number of scenes based on video length (15 seconds: 3 scenes, 30 seconds: 5, 60 seconds: 10), not on the number of photos: 0.385 USD if the scene's narration is 5 seconds or less, 0.77 USD if longer (up to 10 seconds). Step 4 shows the calculation, for example "5 scenes × 0.385 USD = 1.93 USD, up to 5 × 0.77 USD = 3.85 USD". Only clips that are actually created are charged; this doesn't include the cost of writing the script, the voice and photo analysis. If your balance isn't enough for all the clips, step 4 warns you in advance and offers a Use still images button; while running, the app creates no clips at all rather than doing half of them.

The app never replaces a clip with a still image (slow zoom) without telling you: the card in Queue and the details in History state which scenes used still images and why. Rejected or failed scenes are not charged. A video that stopped midway, or finished with some clips still missing, has a Continue button (in Queue or History): it reuses every clip already created (including files already on your computer) and the saved task IDs, without resubmitting or rewriting the script; it only submits scenes that never had a clip, with an estimate. Each video can have only one resume run at a time. When cost reconciliation finds a clip has finished, the app fetches it automatically (without submitting a new one). If a submission didn't return a task ID (network error), the Clip task ID for scene N field lets you paste the ID (a string of digits) from the service's website; the app checks the ID before saving it.

Click Check key in the OpenRouter tab (or Check key in the setup assistant). The app shows your account balance and the key's remaining limit, and warns you when the balance is low.

If the app says it couldn't read the balance: OpenRouter only lets management keys see the account balance, so check your balance on openrouter.ai before creating videos.

• Set a Credit limit on the key directly on OpenRouter (about 2 USD is reasonable). When the key reaches its limit, the app shows an error and stops, spending nothing more.
• Each video has a cost cap (see the question below). A video never goes over its cap.
• Check the Estimate line in step 4 before clicking Add to queue. It's a preliminary figure based on current OpenRouter prices.
• By default the app uses cheap models; choosing AI video (more expensive) or RunningHub AI clips costs more.

The maximum amount one video can spend, in USD, depending on the video type:

• Your photos/videos (no AI clips): 0.10 USD. If there are videos and you choose to fill scenes with AI images, the AI image cap applies; with AI video, the AI video cap applies.
• AI image: 0.30 USD.
• AI video, AI motion clips (RunningHub) or presenter: 6.00 USD (enough for about 8–15 RunningHub photo clips).

A video that mixes several types uses the cap of the most expensive type. Change these three amounts under System settings → Cost cap per video.

The app checks the cap three times: in step 4 (if the estimate exceeds the cap, you can't add it to the queue yet; the app suggests shortening the video, using AI images instead of AI video, using your photos/videos, or turning on Raise the cap this time), right before the video starts, and after the script is ready, before the most expensive step (images, AI video, clips). While running, if the actual cost hits the cap, the video stops completely; the parts already created are kept, and History records the amount spent. A paid step already running still finishes.

The USD → VND rate is for reference only and is fetched online once a day (the Vietcombank selling rate, with open.er-api.com as a backup). Without internet, the app uses the most recent rate it fetched; if it has never fetched one, it uses 26,000 ₫/USD for now. Caps and all comparisons are always in USD.

These are paid steps whose actual cost isn't known yet, for example when the network fails right after a request is sent, or the app closes while waiting. Their estimated cost is still counted in the video's cost. Each time you open the app, it asks OpenRouter and RunningHub again for the actual cost; click Reconcile now on the Queue or History page to ask right away.

In the Documents\bos-video-ai\output folder of the Windows account you are using. Click Open video folder (on a finished video's card in Queue, or in the video details in History) to open this folder quickly.

The History page lists every video as a grid of cards, including videos that are Creating, To review (with an Open to review button), Failed, Stopped or Interrupted (the app closed while creating; videos with an approved script have a Continue button). You can filter by status and sort. Each card has View (opens the details: settings, scenes, narration, cost), Download and Delete buttons (the app asks for confirmation; click Delete to delete; this can't be undone). Finished product review videos also have Try another idea to make a video from the next idea. The details page has Download video, Open video folder and Back to the list.

Yes. Each video shows its actual cost in USD with a reference VND amount, plus a reconciling part if some steps still have an unknown actual cost. The details page (the View button) shows the cost by part (script, AI images, video/clips, voice), the video's cap and the reason it stopped (hit the cap, estimate over the cap, provider error, interrupted). Videos created before the app recorded costs show "—".

No. Reloading the page (F5) keeps you signed in. The app does not store your password or sign-in token on your computer.

Your sign-in session lives only in the app's memory. When you close the app, the session is gone, so you need to sign in again next time.

You are also signed out when the session expires or your account signs in somewhere else. In that case (with the app still open), videos being created and drafts are kept for your next sign-in.

It's your name in your BOS account (the employee name your business entered on BOS). If BOS has no name, the app shows your username. Click the name to see the business you're using and your role. To change your name, ask the business owner to edit it in the employee section on BOS; the app shows the new name the next time you sign in.

Click your name at the bottom of the left sidebar (below "Report a bug / support"), then click Sign out. This menu also has a language switch (Tiếng Việt / English) and Switch business when your account belongs to 2 or more businesses.

Go to the app's home and support page: bosvideo.net. The Report an issue / support item at the bottom of the left sidebar and the About section at the bottom of the Help & About page both open this page. When reporting a bug, include a screenshot, the steps you took and the Video ID (in the video details on the History page) if the bug involves a specific video.

The Edge TTS voice uses Microsoft's service over the internet, so it needs a stable connection. Check your network connection and try again, or use Preview voice to test before creating a video.

The key is wrong or has been disabled. Create a new key on openrouter.ai, paste it again into System settings (without spaces or accented characters), click Check key, then Save settings.

• "Insufficient OpenRouter balance": add more credit on openrouter.ai.
• "OpenRouter key has reached its spending limit": raise the key's Credit limit on openrouter.ai.
• "Insufficient RunningHub balance": add funds to your RunningHub account.

AI motion clips use RunningHub's Model API, which only works with Enterprise-Shared keys. Get a key of the right type in your RunningHub account and enter it again in System settings.

The selected model belongs to a direct provider (DashScope, ARK, OpenAI, Kling...) that has no key yet. Choose an OpenRouter model, or enter that provider's key under System settings → Advanced.

Yes. The app detects a low-end computer, meaning one with 8 GB of RAM or less or a CPU with 4 threads or fewer, and switches to a lightweight mode:

• "Video quality" set to Automatic exports 720p. Videos are still sharp enough for TikTok, Reels and Zalo, while being created much faster and with far less stutter.
• The app creates only 1 video at a time.
• While creating a video, the app doesn't draw the subtitle preview.

To export 1080p, choose 1080p under System settings → Video quality; your computer will be slower. While creating videos, close Chrome and Zalo if you can. System check → Computer specification shows how much RAM and CPU your computer has.

Rendering one video needs about 1.5 GB of free memory, and the final video assembly step alone needs about 1 GB. The app checks before starting and before assembly; if there isn't enough, the video stops instead of freezing your computer. Close some open apps (Chrome, Zalo…), then click Continue (or Retry) in Queue. Finished parts (voice, images, AI clips) are kept, so you don't pay again.