Shorts Factory / ENGINEGet source kit
SOURCE KIT / V1.0.0-RC1 / SELF-HOSTED

Own the source.
Operate the pipeline.

A source-available video production engine for technical operators. Run a durable coordinator on Linux/Docker, media processing on Windows, and Google Cloud models on accounts you control.

Get Source Kit from $149 ↗Explore the architecture

One-time source license. Compute, storage and provider usage are separate buyer costs.

EXECUTION MODELHYBRID / v1.0
SF / ENGINECONTROL
PLANEBuyer-hosted · Source-available
GCP / VEO + TTS
DOCKER / CORE
WINDOWS / MEDIA
STATE IS DURABLEOUTPUT IS GATED
SYSTEM BOUNDARIESPOSTGRESQL LEDGERREDIS QUEUEVERTEX AI VEOPULSAR ALIGNMENTS3-COMPATIBLE
01 / SYSTEM ARCHITECTURE

Five boundaries.
One accountable flow.

Explore each execution layer. This is an architecture map, not a live status dashboard or a promise of unattended operation.

EXECUTION TOPOLOGYSELECT A NODE TO INSPECT
INSPECTED NODE01 / INPUT · Buyer-managed n8n

Import nine sanitized workflows into your own n8n instance. Webhooks and schedules remain under your control.

BUYER NETWORKEXTERNAL PROVIDER BOUNDARYDURABLE STATE
02 / OPERATING CONTRACT

Specific about the machinery.
Specific about the limits.

Provider calls, alignment and delivery are separate contracts, each with its own evidence and failure boundary.

A / GENERATE

Google Cloud is the model boundary.

Vertex AI Veo creates scene clips from prompts and directional START/END frames. Cloud TTS supplies narration; Chirp validation is auxiliary and may be unavailable without making the core narration fail.

BUYER-OWNED GCP BILLING + QUOTAS
B / RESUME

Provider state is durable, not magical.

The ledger records stable request identities and Vertex long-running-operation IDs for polling after a restart. If Vertex accepts a create but the process stops before its ID is persisted, recovery fails closed rather than creating another clip.

NO DUPLICATE CREATE ON KNOWN RUNNING LRO
C / ALIGN

Pulsar decides publishability.

The Windows worker uses Python, faster-whisper and local CMU pronunciation data for word-level timing. A deterministic fallback can make a diagnostic render, but only verified PULSAR_ALIGNMENT_COMPLETED evidence can promote approved media.

DEGRADED ≠ CANONICAL
D / DELIVER

Quality precedes dispatch.

Veo creative generation is bounded to two total attempts per scene, including any RAI-filtered first attempt. Scene QA and Final QA follow; zero critical failures and canonical alignment are required before approved media can reach Telegram.

TELEGRAM_ONLY PROFILE IN v1.0
Render contractWindows + FFmpeg are supported. The current canonical encode path uses libx264; NVENC is not a default or guaranteed acceleration path.
03 / DEPLOYMENT PREREQUISITES

What ships. What you bring.

No bundled n8n server, storage server, cloud credit or model weights. The documented v1.0 topology requires a Windows media worker.

01 / INCLUDED

Source Kit

Versioned, inspectable components and setup evidence.

  • Coordinator and Windows worker source
  • Nine sanitized n8n workflow JSON files
  • Pulsar source and pinned Windows dependencies
  • PostgreSQL migrations and operation ledger
  • Health checks, offline validation and SHA-256 manifest
  • Hybrid deployment and recovery runbooks
02 / BUYER-PROVIDED

Your infrastructure

Capacity, credentials, provider access and operating costs.

  • Windows x64 media host; NVIDIA hardware is optional, not a codec guarantee
  • Linux/Docker host and buyer-managed n8n
  • PostgreSQL, Redis and S3-compatible storage
  • Google Cloud billing and Vertex AI Veo / Cloud TTS access
  • Configured GCS output prefix for Veo transport
  • Telegram bot for the canonical telegram_only delivery profile
INSTALLATION BOUNDARY

Import workflows into your own n8n. Connect your own S3-compatible service. Veo output may transit your configured Google Cloud Storage prefix before durable S3 storage; provider artifacts and cloud charges are not included in the license.

04 / SOURCE LICENSING

Choose the operating rights
your team actually needs.

One-time license price, separate from infrastructure and third-party usage. Review license terms before checkout.

01 / DEVELOPER
SINGLE OPERATOR / INTERNAL PIPELINES

Developer

For your own brands, channels and internal productions.

$149USD / one-time
  • One operator and one hybrid production deployment
  • Internal pipelines for your own brands and channels
  • Core, Windows worker and Pulsar source
  • Nine n8n imports, runbooks and notices
Buy Developer License
02 / AGENCY
UNLIMITED COMMERCIAL RIGHTS FOR CLIENT PROJECTS

Agency

For client-facing teams operating licensed production nodes.

$399USD / one-time
  • Your authorized team on up to three hybrid deployments
  • Unlimited finished-video rights for client projects
  • Everything in the Developer kit
  • Multi-client operational guidance
Buy Agency License

Agency rights cover client media outputs, not sublicensing the engine. Update and support terms follow the purchase order and license.

05 / BEFORE YOU CHECK OUT

Clear answers.
No hidden stack.

If your setup differs from the documented hybrid topology, confirm compatibility before purchase.

A Linux/Docker coordinator host, a Windows x64 media worker, PostgreSQL, Redis, S3-compatible storage, your own n8n instance and a billed Google Cloud project with Vertex AI Veo and Cloud TTS access. The canonical telegram_only delivery profile also needs your Telegram bot. Check provider quotas and the hybrid runbook before buying.

OWN THE PIPELINE

Keep the source.
Keep operational control.

Need to validate an agency deployment or ask about licensing before purchase?

Contact commercial support Review licenses