nolanxai

NolanX

Open-source long-runtime multi-modal agent infrastructure for filmmaking.

NolanX is the system behind nolanx.ai. It is designed to push video models far beyond their native single-generation limits and keep long creative runs coherent across language, image, audio, and video.

With the right model stack, NolanX can turn short-form generators such as Seedance 2.0, Kling 03, Google Veo 3.1 Lite, LTX 2.3, PixVerse v6, Wan 2.7, and others into much longer production pipelines: 5 minutes, 30 minutes, even 1 hour.

It also keeps consistency across large agent workloads: story logic, screenplay, script breakdown, character views, scenes, props, sound, auditions, world assets, timeline editing, storyboard structure, dialogue, performance detail, and directorial reasoning.

Why this exists

There are only two of us maintaining reelmind.ai and nolanx.ai.

I’m a 10-year entrepreneur, now living as an AI nomad between Serbia, Europe, and North Africa.

One day I told my partner Keii:

If you can build something where one click, just one click, generates a movie like Nezha 3 or Interstellar 2, I’ll call you Dad.

Eventually, I built that technology myself.

I believe almost every serious AI film builder will end up extending models this way. And I believe many AI companies will too.

That is why I chose to open-source NolanX.

Sorry, Netflix. When I finally found the right disruptive slogan, I thought of you.

Have fun.

Architecture

This repo is trimmed to the NolanX path. Community, membership, admin, and credit logic are out of scope here.

Use NolanX

1. Use nolanx.ai directly with membership or credit packs Just open nolanx.ai.

2. Use the hosted API route if you want NolanX runtime behavior without wiring every provider yourself

   • Text: OpenRouter
   • Image: FAL
   • Video: ReelMind, our own API platform with a 5% markup
   • Optional R2: Cloudflare R2

3. Fully self-host it locally or privately Run the full stack yourself: web, API, and agent runtime.

   • Configure apps/agent/config.toml first
   • Use root .env only for local host/port overrides
   • Use /nolanx Runtime Keys when you want to edit runtime overlays
   • Replace providers, priorities, and storage paths as needed

Quick start

Clone and start:

git clone https://github.com/nolanx-ai/nolanx.ai.git && cd nolanx.ai && ./dev.sh

Or, if you already cloned the repo:

cd nolanx.ai
./dev.sh

This starts:

• web: http://localhost:3000
• api: http://localhost:8080
• agent: http://127.0.0.1:52178

The browser opens http://localhost:3000/nolanx automatically after startup.

Development

pnpm --dir apps/web build
pnpm --dir apps/api build
cd apps/agent && .venv/bin/python -m py_compile main.py

Open-source runtime model

The stack is layered by capability.

Text

• key: OPENROUTER_API_KEY
• default model: google/gemini-3.5-flash
• responsibility: chat, planning, screenplay generation, routing, general reasoning
• link: OpenRouter
• self-host note: this is the first provider family you swap or extend

Image

• key: IMAGE_API_KEY
• default model: openai/gpt-image-2
• responsibility: image generation, image editing, keyframes, visual references
• link: FAL
• self-host note: replace the image endpoint/provider with your preferred image stack

Video

• key: VIDEO_API_KEY
• default model: dreamina-seedance-2-0-260128
• default endpoint: https://nestapi.reelmind.ai/external-api/video/generate
• responsibility: shot generation
• link: ReelMind
• routing note: this is our own API platform and this routed path includes a 5% markup
• self-host note: replace this layer with your preferred video provider or internal endpoint

Optional R2

• keys: R2_ACCOUNT_ID, R2_ACCESS_KEY_ID, R2_SECRET_ACCESS_KEY, R2_BUCKET_NAME, R2_PUBLIC_URL
• responsibility: uploaded references, generated assets, continuity-friendly persistent storage
• link: Cloudflare R2
• note: NolanX still runs without it, but storage-backed flows become more limited

What happens if a key is missing

• no OPENROUTER_API_KEY: chat, planning, and script generation do not run
• no IMAGE_API_KEY: image generation and image editing do not run
• no VIDEO_API_KEY: video generation does not run
• no R2_*: NolanX still runs, but enhanced persistence / upload / continuity flows stay disabled

Configure runtime keys

• Open /nolanx and click Runtime Keys for local overlays
• apps/agent/config.toml is the primary self-host configuration file
• apps/api/data/runtime-config.json stores UI-edited runtime values

Local data

Local data lives under apps/api/data:

• nolanx-db.json
• runtime-config.json

Credits

NolanX was inspired, in a small way, by the Jaaz project about a year ago.

Over that year we kept breaking through hard technical problems, modifying LangChain where needed, and learning from strong ideas in OpenClaw, Harness, Hermes, long-memory systems, and long-runtime agent design.

Thanks to the open-source community for the work that made this path possible.

This is our contribution back.

WhatsApp Group

Scan this QR code using the WhatsApp camera to join this group.

Star History