Design the shot. Run the workflow. Keep the results together.
Director's Console brings cinematic prompt design, reference-driven image and video workflows, and project media management into one workspace for filmmakers and AI artists. Author a shot in Cinema Prompt Engineering, render it through ComfyUI in Storyboard, and review the results in Gallery.
Project Eliot
Existing panel overview: per-shot panels, parameters, reference inputs, and node selection are shown; the latest controls are documented in the Storyboard guide.
- Why Director's Console
- Cinema Prompt Engineering
- Storyboard
- Gallery
- Installation and Quick Start
- AI LLM Provider Setup
- Architecture
- Documentation
- Changelog
- Security & Deployment
- Cinematic defaults — Start from grounded camera, lens, lighting, movement, and style choices.
- Image-aware enhancement — Start from text, an attached image, or both; use a vision-capable LLM when images are attached.
- Workflow-based generation — Use ComfyUI workflows for generation, editing, upscaling, and video.
- Per-shot control — Keep panel-specific parameters, versions, notes, and history together.
- Parallel rendering — Select multiple ComfyUI nodes and follow progress for each job.
- Project media organization — Browse, rate, tag, rename, compare, trash, and hand off project files.
CPE is the text and configuration side of the application. Select Live-Action Cinema or Animation, use a preset or configure fields directly, and choose a target prompt profile. The + picker, drag-and-drop composer, and image previews support text-only, image-only, or text-plus-image enhancement.
Generate creates rule-based text from the current configuration. Enhance with AI sends the selected LLM the attached pixels when applicable, plus the full current configuration and selected preset/style context; choose Describe image, Use as reference, or Use as starting image, with model-specific variants where supported. CPE does not render.
Film preset browser beside the configuration experience: camera, lens, lighting, and preset choices guide the brief. See the Cinema guide for current controls and boundaries.
Guides: Cinema Prompt Engineering · Cinema presets · Cinematography reference · Enhancement profiles
Storyboard is the execution workspace. Import or select a ComfyUI workflow, expose its supported controls, assign per-shot parameters, add image/video inputs where the workflow allows them, and send the workflow directly to one or parallel ComfyUI render nodes.
Actual image/reference generation, editing, upscaling, and video generation depend on the selected workflow, installed weights, and custom nodes. The frontend uses ComfyUI REST and WebSocket APIs for execution; Storyboard Enhance remains metadata-only for workflow media and does not render or send media pixels to an LLM.
Guide: Storyboard · Multi-node generation
Gallery is the project media tab beside Cinema and Storyboard. Use grid, list, or borderless masonry views, with timeline grouping for chronological browsing. Filter, compare, rename, rate, tag, search metadata, move files, and use trash/restore without leaving the project.
Gallery can send a reference to an exposed Storyboard input or restore compatible workflow metadata from a PNG. That handoff prepares a workflow input; the selected workflow still determines what can be submitted.
Guide: Gallery
- Open Cinema Prompt Engineering, choose a discipline and preset, then tune the brief.
- Use Generate for deterministic rule-based text, or attach an image and choose Enhance with AI when a configured vision-capable LLM is available.
- Send the chosen prompt to Storyboard, select a workflow, and confirm the panel's exposed parameters and image inputs.
- Choose one or more render nodes and generate. Outputs remain in the panel history and can be reviewed or organized in Gallery.
- Python 3.11+
- Node.js 22.13+ (22.x) or 24+, with npm
- Git
- Windows, macOS, or Linux
- For rendering: a running ComfyUI node with the selected workflow's models and custom nodes. Prompt authoring does not require a render node.
git clone https://github.com/NickPittas/DirectorsConsole.git
cd DirectorsConsole
python start.py --setup
python start.py--setup provisions the launcher's per-service environments and exits; it does not boot all services. The launcher then provides:
| Service | Address | Role |
|---|---|---|
| Frontend | http://localhost:5173 | Director's Console UI |
| CPE backend | http://localhost:9800 | Rules and prompt API |
| Orchestrator | http://localhost:9820 | Project, file, Gallery, and job-group operations |
Optional flags: python start.py --no-browser, python start.py --no-orchestrator, and python start.py --no-frontend.
For manual setup, development commands, and regression checks, see Contributing. After launch, open the frontend and configure a project destination before saving generated media. Add ComfyUI node URLs in Storyboard only when you are ready to render. Provider configuration is similarly optional unless you use Enhance with AI.
Provider setup is optional for rule-based Cinema prompts and required only for Enhance with AI. Compatible local servers and existing OpenAI Codex OAuth are supported where configured. A model catalog or connection does not prove account access or vision availability.
Use the Google AI (Gemini) provider with a Google AI Studio API key. This is distinct from Antigravity OAuth; never put a real key in source, logs, or frontend VITE_* variables. See the Google AI setup details.
Antigravity uses authorized OAuth application credentials in the private backend CinemaPromptEngineering/.env, not frontend variables. Never commit client secrets or tokens. See the Antigravity configuration details.
See the complete provider guide for compatible endpoints, local servers, credentials, and verification boundaries.
- Frontend: calls selected ComfyUI nodes directly through REST/WebSocket for workflow generation.
- CPE API: provides cinematography rules, configuration, and optional LLM enhancement.
- Orchestrator: handles project/file operations, Gallery services, backend status, and job groups.
See Multi-node generation and the architecture details in the guides.
| Topic | Guide |
|---|---|
| Cinema | docs/cinema.md |
| Storyboard | docs/storyboard.md |
| Gallery | docs/gallery.md |
| Providers | docs/providers.md |
| Presets | docs/cinema-presets.md |
| Technical reference | docs/cinematography-reference.md |
| Enhancement profiles | Documentation/PROMPT_ENHANCEMENT_PROFILES.md |
| Multi-node rendering | docs/MULTI_NODE_GENERATION.md |
| Development | docs/contributing.md |
| Recovery | docs/session-recovery.md |
- Added image attachments, description/reference/starting-image modes, and model-specific variants for AI prompt enhancement.
- Preserved complete configuration and preset/style context; improved compatible endpoints, model discovery, OAuth, and provider responses.
- Reorganized this README around the three tabs, with focused guides for panels, providers, presets, and technical reference.
Offline implementation checks passed. Live provider, browser, and render verification remains pending; see the verification details.
Added workflow categories, expanded image guides, Gallery layout work, and translated-DOM control hardening.
See CHANGELOG.md for the full history.
Director's Console has no application-wide authentication or authorization. Use it only on a trusted, private, access-controlled network; it is not a hardened public service and has not undergone a formal security audit. Protect provider credentials, browser profiles, backend data, workflows, media, and connected ComfyUI nodes. Treat project files and workflows as untrusted input.
See the provider security details and the contributing guide for operational boundaries.
Released under the MIT License.