Director's Studio v3.6.7

DIRECTOR'S STUDIO

A PRE-PRODUCTION CONSOLE FOR AI VIDEO

STEP 1 / 3 — WELCOME

Your data stays in your browser. Nothing is uploaded to us.
You bring your own LLM — cloud (Anthropic / OpenRouter / MiniMax) or local (llama.cpp / Ollama / LM Studio).
API keys are session-only — wiped when you close this tab. You'll re-enter them next session.

STEP 2 / 3 — PICK YOUR BACKEND

WHERE'S YOUR LLM?

☁️

CLOUD

Anthropic, OpenRouter, or MiniMax. Best quality, pay-per-token.

CONFIGURE →

🖥️

LOCAL

llama.cpp · Ollama · LM Studio · vLLM. Free, private, your hardware.

CONFIGURE →

STEP 2 / 3 — CLOUD BACKEND

CONFIGURE CLOUD PROVIDER

PROVIDER

Endpoint:

API KEY

Stored in sessionStorage — wiped on tab close.

MODEL

Pick a shipping model. You can edit this later in Settings.

STEP 2 / 3 — LOCAL BACKEND

CONFIGURE LOCAL LLM

ENDPOINT URL

llama.cpp: http://localhost:8080/v1/chat/completions · Ollama: http://localhost:11434/v1/chat/completions · LM Studio: http://localhost:1234/v1/chat/completions

REQUIRES API KEY

MODEL NAME

Local model names depend on your server's config. Free-text.

⚠ Local servers must allow CORS. Troubleshoot ↗

STEP 3 / 3 — READY

✓ CONNECTED

By default, all 3 stages will use this backend.

💡 VISION CAPABILITY — reference images require a vision-capable model. Director's Studio supports vision on:

Anthropic Claude (Sonnet/Opus 4.x — full support)
OpenAI GPT-5.x / GPT-4o (full support)
Google Gemini Pro / Flash (full support)
MiniMax M2 / M2.7 (auto-routes to /anthropic/v1/messages vision endpoint)
Local LLMs — depends on model (Gemma 4, LLaVA, Qwen-VL, Pixtral, etc.)

Text-only providers (e.g. DeepSeek chat) will ignore attached refs.

▸ POWER USER? USE DIFFERENT MODELS PER STAGE

You can configure more presets and fast/smart-tier routing now, or do it later in Settings.

PROJECT: untitled Click name to rename · 💾 saves · 📂 opens library

FRESH ROOTS:

starting concepts

FEELING / THEME / MOOD / LYRIC

✨

Drop a feeling / theme / mood / lyric above.
Get back 5 distinct concept directions — lock any, deepen any with AI.

NO IDEA → 5 IDEAS · LOCK + SEND TO BRIEF CONCIERGE

🔒 LOCKED CONCEPTS (0 / 5)

LOCK IN UP TO 5 CONCEPTS
FROM THE TREE OR AI SPARK
THEN SEND TO BRIEF CONCIERGE

🎬 INPUT

YOUR IDEA (any depth)

SONG LYRICS (optional — triggers music-video mode)

REFERENCE NOTES (optional)

STYLE / MOOD REFERENCES (optional)

📎

Drop / paste / click — up to 4 refs

PNG · JPG · WEBP · CLICK ROLE BADGE TO CYCLE STYLE / CHAR-1/2/3 / SCENE / VIBE-1/2

🎲 Options for every field

→ Forge new character

📋 BOARD-READY BRIEF

📭

Drop your idea on the left and click Generate Brief.
The concierge fills every field in one pass — you tweak from there.

DRAFT-FIRST · NEVER ASKS · 🟡 MARKS DEFAULTS

🎬 CONCEPT

IDEA / BRIEF

🎭 CHARACTER

DESCRIPTION

🎚 PRODUCTION

TONE

DURATION (SEC)

ASPECT RATIO

PANEL COUNT

STORYBOARD LAYOUT

CHARACTER SHEET LAYOUT

VISUAL STYLE

PROJECT TITLE

🖼️ IMAGE GEN TOOL

IMAGE GENERATOR

🎯 VIDEO GEN TOOL

VIDEO GENERATOR

🎵 AUDIO

SONG LYRICS (auto-IPA per panel)

🖼 REFERENCE IMAGES

📎

Drop / paste / click — refs go to AI vision (if model supports it)

PNG · JPG · WEBP · CLICK ROLE BADGE TO CYCLE STYLE/CHAR1/CHAR2/CHAR3

🔁 REFINE

🎬

Fill the brief, drop reference images, paste lyrics. Hit GENERATE STORYBOARD.
Get back a storyboard preview, character sheet prompts, IPA sync block, and per-panel video prompts — all copy-paste ready.

⌘+ENTER to generate · CLICK ANY PANEL TO EDIT · NO IMAGE/VIDEO API CALLS

📺 Composite Preview · Step 3 Bake-in Drop your generated storyboard image, then click 📝 Generate Bake-in Prompt

📥

Drag & drop your generated storyboard image here

Or click to browse · paste from clipboard · PNG / JPG / WebP

First generate above + paste the storyboard prompt into your image gen, then drop the result here.

🎭 INPUT

CHARACTER LABEL (literal — used as the name)

REFERENCE PHOTO (optional)

📷

Drop / paste / click — one photo

PNG · JPG · WEBP · DESCRIBES WHAT YOU SEE

TYPE

DETAILS (any depth)

🪪 CHARACTER

🪪

Drop a character label and any details on the left.
Get back a paste-ready one-liner plus a structured breakdown.

FACE · HAIR · BUILD · AGE · WARDROBE · TRAIT · ANTI-BIAS

📂 PROJECT LIBRARY

💡 RECENTS store a light copy in your browser (no reference images / composite). For full archival use 💾 to download the .dsp.json. Auto-trims to 10 entries.

📂 PROJECT LIBRARY

ℹ About token usage

llama.cpp

Ollama

LM Studio