White Label AI Video Generator Features: Script to Localized Export
Generating a clip is the easy half. What you buy here is everything a business wraps around generation: client workspaces, versioned scripts, rendering on worker queues, translation as its own stage, plan limits checked on the server, a metered API and an eighteen-tab command center. You own the source, and no per-video or per-seat fee comes back to us.
Request a Live Demo →See the CostFeature Set by Role
This is the platform underneath our AI video clones, and the live demo hands you a login for each of its four roles: viewer, creator, manager and platform admin. Viewers simply watch finished output and read notifications; the three cards below cover the roles that do the work. Each of them operates inside a workspace, the unit that lets one deployment host many teams or clients side by side.
Creator
Where a script turns into finished, exportable video.
- Script editor that keeps each revision as an immutable snapshot, with rollback
- Avatar and voice picked per project from searchable catalogs
- Rendering handed to a worker queue, with states from pending to completed or failed
- Reusable public and workspace templates, composed scene by scene
- Localization jobs where language, voice and subtitles are chosen each time
- Exports from format and resolution presets, bounded by the plan
Workspace Manager
Governing a team, its plan and its integrations.
- Invitations with owner, editor or viewer scope, and pending invites you can cancel
- Plan, billing cycle and usage visible for the workspace
- API keys metered per endpoint and per day
- Webhooks with delivery logs, retries and rotating secrets
- Media library organized by folders and tags, with bulk actions
- Workspace analytics exported as CSV or JSON
Platform Admin
The operator's eighteen-tab view across the entire deployment.
- Suspend, re-own, re-plan or delete any workspace
- Suspend users, reset passwords or impersonate one for support
- Archive, restore or move projects between workspaces
- End a single session or every session a user holds
- Nine integration sections edited at runtime, each with Save and Test
- A platform-wide audit trail of who did what to which resource
Localization is stored as its own record with a dedicated worker, not as a second pass over the original render. A Spanish version that fails can therefore be retried by itself, and the English master is never touched.
Rented AI Video Platform vs One-Time Script vs Miracuves Platform
Three common routes to an AI video product that carries your brand, and where each one tends to pinch once paying customers arrive.
| What decides it | Miracuves platform | Rented or hosted platform | One-time script |
|---|---|---|---|
| Who holds the source | You, in full, free to change and redeploy | The vendor; you rent access | You, though often only a thin wrapper |
| How you pay for the software | $3,699 once, no per-video or per-seat fee | Recurring seat or credit fees | Usually a single payment |
| Generation cost | Billed to you by the AI providers you connect | Usually folded into the vendor's credits | Billed by whichever API it wraps |
| Workspaces and roles | Workspace tenancy with owner, editor and viewer scopes | Team accounts, within the vendor's limits | Often missing |
| Localization | A separate stage with its own worker and status | Whatever the vendor exposes | Typically a fresh render, if present at all |
| Plans and metering | Server-side plan limits, export tiers, metered API keys | Resale margin on the vendor's pricing | Billing usually left for you to build |
| Admin and audit | 18-tab command center with platform-wide audit logs | Reseller panel with limited controls | Rarely more than a settings page |
| Customer data | In your PostgreSQL database, on your infrastructure | On the vendor's servers | On your server |
| If you stop paying anyone | Nothing changes, the code is yours | Access usually ends | Nothing changes |
A rented platform does bundle something real: hosting and model access in a single bill. Here the servers and the provider accounts are yours to run, so the fair comparison is a one-time software cost against a recurring fee, not the total cost of launch. The full breakdown sits on the Development Cost page.
The Technology Behind the Features
One monorepo in four packages (frontend, backend, shared types and utilities), so the web app and the API agree on their contracts. The front end is React and TypeScript bundled with Vite. The API is Express with Zod validation, versioned REST under /api/v1, GraphQL with complexity caps and a Swagger surface your developers can browse. Data lives in PostgreSQL through Drizzle across ten ordered migrations. Generation, translation, export and webhook jobs run in four queue domains on Bull and Redis, spreading across instances, with an inline mode kept for staging and small installs. The interface speaks English, Spanish, French, German, Hindi, Arabic, Chinese and Portuguese, with 156 keys in each locale, and changes language without a reload.
From Blank Script to Measured Video
From outside it looks like one button. Inside, every video passes through six stages that each keep their own state, and that separation is what makes retries, progress tracking and diagnosis possible.
Open a workspace
Signing up creates a workspace and its owner membership in a single transaction, and TOTP two-factor can be switched on straight away. Every asset, template, webhook and subscription that follows belongs to that workspace.
Write and revise
A project holds the brief and its lifecycle status; the script holds the words. Edits are saved as immutable snapshots, so returning to the draft a client preferred three rounds ago is a rollback, not a rebuild.
Render on a queue
Before a job is accepted, the service layer confirms the plan still has room. The job then goes to the generation worker and moves from pending to processing to completed or failed, with each attempt logged against its provider and settings.
Localize
Choose a target language, voice and subtitle options and the job goes to a worker of its own. Its status is tracked apart from the master video, and notifications and webhooks fire when it finishes.
Export and publish
Format and resolution presets feed an export queue with history and a download lifecycle. The plan decides which resolutions are allowed, and distribution endpoints support automated publishing.
Measure
Workspace KPIs, per-video detail and breakdowns by country, device and event give a marketing lead something concrete for the budget meeting, while usage metrics guide your own plan and capacity choices.
Every Feature Earns Its Place
Why each capability belongs in an AI video product you plan to sell, and what an operator gives up without it.
| Capability | How it behaves | Why an operator needs it |
|---|---|---|
| Workspace tenancy | Nearly every business table references a workspace | Separate clients, per-workspace plans and per-workspace API keys run on one install instead of many |
| Server-side plan limits | Checked in the service layer before a job is queued | A tampered browser cannot slip past them, and no provider bills you for a render you cannot charge for |
| Worker queues | Four queue domains, Redis-backed across instances or inline | A slow provider lengthens the queue rather than producing a wall of timeouts, and a failed job can be retried |
| Immutable script versions | Each revision kept as a snapshot | Agencies and training teams update content with an edit, never a reshoot |
| Separate localization stage | Own entity, worker and status per language | One market's failure never blocks another, and translation can carry its own entitlement and price |
| Export entitlements | Resolution and format rules enforced in the export service | An upgrade customers can feel, rather than a button merely hidden in the interface |
| Metered API keys | Usage logged per endpoint per day, per workspace | A developer tier you can price without building a second billing system |
| Webhooks with delivery logs | Retries, rotating secrets and SSRF screening of target URLs | Customers automate around your platform, and a failed delivery is visible instead of silent |
| Runtime provider settings | Nine integration sections stored in the database | Switch AI providers as pricing or quality shifts, with no redeploy |
| Audit logging | Records the actor, the action and the resource | The evidence a procurement or security review asks to see |
What Is Not Included in the Base Package
The workspace model, the render and localization pipelines, plan gating and the command center are built. What follows is not, and several items are configuration you complete before real customers arrive.
- An AI model of our ownMiracuves has none. Generation runs through AI provider accounts you open and pay for, under those providers' terms, so output quality and render cost depend on them. We connect the accounts; you hold the keys.
- Rights to any face or voiceThe software supports custom avatars and validated voice-cloning uploads, yet it grants no right to a likeness. Written consent from each person whose face or voice is used, labeling of synthetic media where your markets require it, and moderation of what customers produce are your duties.
- Native mobile appsNone ship. It is a responsive web platform, suited to production work done at a desk. Native iOS and Android can be scoped as custom development.
- Live billingPlans, entitlements and usage are modeled and a Stripe-compatible flow is integrated. Charging real money still needs your own Stripe account with its webhooks connected.
- Central authorization middlewareWorkspace scoping runs through the product, but resource-level checks are uneven across route families. One central policy guard, backed by tests that attempt cross-tenant access, is the documented fix and an add-on.
- CORS, CSP and secure cookiesAn origin allowlist and a tighter content security policy are set against your domains at deployment. Moving tokens out of browser storage into secure cookies is a documented hardening step.
- Encrypted secrets and hashed API keysProvider secrets are encrypted at rest once you provision an ENCRYPTION_KEY. API keys are stored raw today; hashing them, with plaintext shown once at creation, is a hardening add-on.
- Enterprise single sign-onSocial sign-in connectors are included and configured at runtime. SAML or OIDC for corporate identity providers is extra work.
- Monitoring and alertingHealth and readiness endpoints and structured logs ship. A log sink, queue-lag monitoring and alerts on auth anomalies, webhook failures and worker errors belong in place before high volume.
- Retention and privacy automationPrivacy duties for customer data stay with you as the operator. Retention windows, scheduled purges and PII classification across analytics, audit and event data are scoped as add-ons.
- Scale-out tuningRedis-backed queue mode is a production requirement, not an option. Spreading queues further, tuning indexes to your highest-cardinality filters and capacity instrumentation follow as volume grows.
The documentation files these items under configuration required before production, and we repeat them here so nothing surprises you after purchase. One installation runs one operator brand, while your customers' brands live in their own workspaces. If you will sell to enterprises that test tenant isolation, take the central authorization hardening before they do.
Deciding who should supply your AI video platform?
The Provider page sets rented platforms, script sellers and custom agencies side by side, lists the questions worth asking any of them about source ownership, customer data and provider accounts, and walks through delivery step by step.
Frequently Asked Questions
How does AI video localization work here?
Which AI models does it use, and who pays for rendering?
Can customers make avatars from their own face or voice?
Can one deployment serve several clients?
What happens when a customer hits a plan limit?
Is there a native mobile app?
Explore the White Label AI Video Generator
Try all four roles before you commit
Write a script as the creator, check usage and API keys as the manager, then open the admin command center and read the audit trail your session left behind.
Miracuves is an independent software development company. We are not affiliated with, connected to, sponsored by, or endorsed by HeyGen or any other AI video, avatar or synthetic media service.
“White label AI video generator” describes a category of product, not any one company. Brand names appear elsewhere on this site only to describe the kind of platform being built and the terms buyers search for.
We supply software, not rights to any likeness or any AI model. Opening and paying for AI provider accounts and staying within their terms, obtaining written consent from every person whose face or voice is used for a custom avatar or voice, labeling AI-generated media where the law requires it, privacy notices, and moderating the content your customers create are your responsibility. We do not advise on any of it.
The entire design and codebase is built by our own team. The product contains no code, design, graphics, or content originating from any third-party AI video website or application. All third-party names and marks belong to their respective owners.
