Armoured Duel in the Stone Arena
Reference input · 3 images
Output
15s · 720p · real output
Generate images and video straight from your browser – no install, no code, no studio booking. Or call the same engine from your own product with an API key. Either way the finishing layer comes attached: re-frame to 9:16, burn in subtitles, dub to another language, erase a logo, restore to 4K. Priced in Ringgit.
Real outputs – click any thumbnail to enlarge it.
01 / Demos
Real outputs, with the input and the prompt beside them. Pick a tool on the left – every tab on this page is wired to samples we can actually show, and the tools we cannot show yet are listed as such instead of dressed up.
2 playable cases · the published reference images, the prompt, and the finished video
10 presenter clips · one product photo, three tones, three subtitle tracks – all generated, none of them filmed
All ten are real model output, trimmed to five seconds from 15-30 second generations, and muted on this page. No actor was booked, no location was hired, and the same product photo is behind all four styles in the first row.
8 text-to-image outputs · Hunyuan 3.0 and the WAND image line
Product three-views · EN
Product three-views · JA
Product three-views · KO
Medical atlas · annotated
Manga page
Photo-real scene
Poster · Chinese type
Traditional paintingOne prompt produced all three product three-views – the same model rendering dimension labels in English, Japanese and Korean. In-image lettering is the part most models still fail at, which is why it is the first thing we show.
4 before / after pairs · drag the handle to compare
1 pair · subject lifted off a busy scene onto a clean plate
Samples generated on Tencent Cloud media AI and reproduced with the permission of an authorised partner. The video cases show their published reference images, prompt and output, unedited.
02 / What is this
Most Malaysian teams do not need another model. They need the last 20 percent: the same video in nine-by-sixteen, in Mandarin, without the supplier logo on it, and small enough to send. That finishing work is the layer almost nobody bundles – and it is what this is built on.
Images, short video, 3D assets and voice from a brief, a photo or a product sheet.
Read a video and pull the summary, the best moments and the scene boundaries out of it.
Erase, enhance, re-frame, subtitle, translate, dub and separate audio tracks.
Compress and encode for the channel you are actually publishing to.
03 / How it runs
Four steps, and you choose how much of them you want to own.
Assets are processed in our Singapore region. We do not claim your data stays inside Malaysia – it does not, and we will not pretend otherwise.
04 / Use it in the browser
Most people do not want an API key. They want a product photo turned into five ad variants before lunch. So the same engine also ships as a plain web app – prompt, recipe, generate, download. No install, no design software, no waiting for someone else.
Preview · generated
Start from “product on white”, “add a Bahasa caption” or “make it vertical” instead of writing a novel to get one decent frame.
Save your logo, colours and product shots once, and keep them consistent across a whole set.
Generate four variants, or a hundred, and keep the ones that work.
9:16 for Reels, 1:1 for the feed, 16:9 for the website – sized on the way out.
No install, no plugin, no design software. It runs in a browser tab on a phone or a laptop, and downloads land in Ringgit-billed credits you can top up yourself.
05 / The catalogue
Pick the outcome you need, not the model behind it. Every capability below is bought by what it produces.
Text-to-image or reference-to-image, any style, from one sentence or a product photo.
Keep the same product, model or character across a whole set of images instead of one lucky frame.
Repaint one area, or extend an image beyond its edges when a channel needs a different crop.
Put a garment on a model, or lift a subject onto a new background, without a studio.
A prompt or a still becomes a short clip, in the aspect ratio you ask for.
Longer pieces that hold the same character and scene across cuts instead of drifting.
Feed in product or character images and direct the shot around them.
Clips that arrive with audio attached, so the first three seconds hold attention.
Drive the camera move from a reference clip instead of describing it and hoping.
Voiceover in the language you sell in, and royalty-free background music.
Auto-reframe a horizontal video for Reels, TikTok and Shorts with the subject kept centred.
Portrait to landscape when the same footage has to sit on a website or a TV.
Automatic subtitles with translation, and the style controls to match your brand.
Transcribe, translate and re-voice in one flow, keeping the original timing.
Erase the old subtitles and logos first, so the new version does not show two languages at once.
Blur, mosaic or remove logos, subtitles, plates and faces from video.
The same on images: text, watermarks and stray objects, cleanly.
Take out what should not be in the frame, including people.
Remove noise and artifacts, lift colour and faces, and push standard footage toward HDR.
Upscale an image or clip, or shrink a file while keeping it looking sharp enough to publish.
Generate a 3D model from a description or a reference image.
Build a world so the background holds steady across every shot of a story.
Capabilities marked Roadmap are not open for early access yet. We would rather show you the gap than imply a date we cannot hold.
06 / Three ways in
Open a tab, type what you want, generate, download. Built for marketers and shop owners, not engineers.
no install – browser onlyOne key, one endpoint shape, generated output back in the response. Batch the boring parts.
{ “model”: “media/video”, “prompt”: “…” }Install a skill so your agent can generate, finish and file media without a human in the middle.
install: bd-media-skill07 / How you are billed
Prepaid and monthly packs that work the same way whether you spend them in the app or through the API – from a small trial to a production quota with a human checking the output.
Higher resolution, sound, motion control, dubbing, priority queue, and extra revisions on managed work.
You send the brief, we return finished files. Quoted per project, and quoted before we start.
Early access members get launch pricing first, and the price we quote you is the price on the invoice.
08 / Questions
No, and we will not pretend it is. We are an authorised Tencent Cloud partner, and the generation runs on Tencent Cloud media AI. What we add is everything around it: the localisation and finishing workflows, RM billing and e-invoice, a Malaysian team, and an API and agent entry point that Tencent does not offer locally.
Because we are onboarding early-access customers first, and we would rather publish a price we have already honoured than a number we guessed. Request access and you get launch pricing before it goes public.
In our Singapore region. We do not claim your data stays inside Malaysia – it does not, and any provider telling you otherwise with this class of service should be checked carefully. If your own compliance rules require something stricter, tell us and we will say honestly whether we can meet it.
Nothing is charged to you for a job that returns no usable output. If you have already paid for credits, the balance is simply not reduced.
Yes, for the business that bought the credits. Tell us at signup if you are producing for a client rather than for yourself, because that changes the account shape, not the price.
No. The web app is the whole point – if you can use Canva, you can use this. The API path is there if you want generation inside your own product, and we will write the first call with you. The managed path needs nothing but a brief.
Same engine, same credits, different door. The app gives you prompts, recipes, your brand kit and export presets with nothing to build. The API gives you the same capabilities as raw calls so you can put them inside your own product. Plenty of customers start in the app and move to the API once a workflow proves itself.
Yes – it is a browser app, so a phone or a tablet works. Long video renders are more comfortable on a laptop, and we will tell you honestly where the phone experience is still rough.
Shorter than a shoot, longer than a search. Image work is near-instant, clips take minutes, and longer or higher-resolution pieces queue. We will give you real timings on your own brief during early access rather than a marketing number here.
Yes – that is most of the reason this exists. Ringgit invoice with SST where it applies, e-invoice compliant, payable by FPX or DuitNow, from a Malaysian company you can call.
We are opening media AI to a limited first group so we can onboard properly instead of quietly. Tell us what you need to produce and we will keep your spot – early members get the first keys and launch pricing.
Prefer to talk? Message us on +60 11-5581 8160 and tell us what you are trying to make. If it is a job we cannot do yet, we will say so on the spot.
Media is one layer. These are the ones that usually sit around it.
Plays with sound.