The studio for novel-to-film adaptation

From manuscript
to motion picture.

One studio for the whole adaptation — manuscript, worldmap, screenplay, render canvas, voice & audiobook. With a per-project AI agent that remembers every page, every shot, every cut.

No credit card·Free tier forever·MCP-native
Nano Banana ProFlux ProRealvisXLVeo 3.1Grok ImagineElevenLabs v3MeshyComfyUIClaudeGeminiModel Context ProtocolNano Banana ProFlux ProRealvisXLVeo 3.1Grok ImagineElevenLabs v3MeshyComfyUIClaudeGeminiModel Context Protocol

Inside the studio

Six surfaces.
One source of truth.

Bible, worldmap, canvas, screenplay, audiobook, and public showcase. Everything you build in one surface shows up in the next — locked, referenced, version-controlled.

cinematicworkshop · The Karakoram Ledger · bible
live·sync 2s ago
BIBLE PREVIEW

A locked production bible.

Characters with face-locked portraits and body-view rigs. Locations with measured blueprints. Props, factions, lore, world rules. Every detail referenced by your renders so nothing drifts between cuts.

60+
MCP tools

Every primitive — chapters, scenes, bible entries, renders — exposed as a workshop_* tool.

5
Image engines

Nano Banana 2, Nano Banana Pro, Flux dev/pro, RealvisXL, plus ComfyUI 4K upscale.

8
Video models

Veo 3.1 and Grok Imagine: image-to-video, first/last frame, extend, text-to-video.

1
Source of truth

Bible refs, worldmap pins, canvas pipelines, voice casts — all in one project.

project-agent · karakoram-ledgersession · 14m32s
[scene-14] drafted 4 beats from chapter 8 ¶3–6[bible] found refs: Renwick, Baltoro camp[canvas] wired image.refs → image.generate(flux.pro)[canvas] wrote image_prompt for beat 1[render] queued · 4K upscale · awaiting result[voice] ElevenLabs/Adam cast: "cold authority"[worldmap] verified Baltoro pin · 47.6km from Skardu[review] scene matches bible · no character drift[agent] ready · awaiting your next instruction

The agent

A collaborator that doesn’t forget.

Every project ships with its own AI agent that holds full memory of everything you’ve built — bible, chapters, scenes, canvases, voice casts, renders that worked.

It writes new chapters, reviews what you wrote, proposes bible entries from your prose, builds canvas pipelines, casts character voices, and queues renders. The same surface a human uses, scoped to the project it lives in.

writereviewpromptimage.generateimage.upscalevideo.veovoice.castmap.verifyscene.draft
workshop_create_scene · MCP callstreamable http
{
  "tool": "workshop_create_scene",
  "args": {
    "chapter_id": "8c2f…",
    "title": "The glacier camp at first light",
    "beats": [
      {
        "subject": "Renwick approaches the tent at -22°C",
        "camera": { "lens_mm": 35, "angle": "low-three-quarter" },
        "environment": "Baltoro Glacier camp, 5800m"
      }
    ]
  }
}

MCP-native

Or call it from your own agent.

Every primitive — chapters, scenes, bible entries, prompts, renders — is also a Model Context Protocol tool. Generate a bearer token, point your client at the URL, and your external agent reads and writes the same project you’re editing in the browser.

Claude CodeClaude DesktopCursorAny MCP client

Capabilities

Everything an adaptation actually needs.
Nothing it doesn’t.

No feature list theatre. These are the six things you’ll use every day — and the slots where your stills, clips, and screenshots will live.

  • Agent UI1280×800

    Per-project AI agent

    Every project gets its own agent with persistent memory of the bible, chapters, scenes, canvases, voice casts. It writes, reviews, generates, casts, queues renders.

  • Bible character render1920×1080

    Image generation

    Nano Banana 2 / Pro, Flux dev/pro, RealvisXL, plus ComfyUI upscale to 4K. References wired per beat. Generation lineage preserved.

  • Scene clip · 8s loop1920×1080 MP4

    Video generation

    Veo 3.1 and Grok Imagine. Image-to-video, first/last frame, extend, and pure text-to-video supported per engine.

  • Audiobook editor1440×900

    Audiobook studio

    Clone any manuscript into an audiobook variant. Per-character voice casting, emotion + SFX cues, ElevenLabs v3-ready scripts.

  • Meshy GLB · turntable1080×1080 WebM

    3D characters & props

    Image-to-3D via Meshy. Body-view rigs (front · back · 3/4) baked into the bible entry for clean GLB generation.

  • Claude Code session1440×900

    MCP-native

    60+ workshop_* tools via Streamable HTTP MCP. Per-user bearer tokens with revoke + expiry. Same row-level security as the UI.

Built for

Whoever you are at the start of the day.

Four jobs. One studio. The same primitives — bible, worldmap, canvas, voice — show up shaped to whatever you came here to do.

For novelists

Take your novel from page to screen — without losing the book.

Keep writing the novel in long-form prose while the agent quietly builds the bible underneath. When you're ready, adapt the chapters you want to see filmed. The two surfaces stay in sync through the bible and worldmap, so a character described in chapter 7 carries the same face and the same coordinates into the screenplay for scene 14.

  • Manuscript editor with chapter-level autosave + revision history
  • Bible suggester that reads your prose and proposes entries
  • Worldmap auto-pins new named places with verified distances
  • Auto-adapter suggests film scenes from finished chapters
  • Export to DOCX, Markdown, or Fountain in one click
Manuscript + bible suggest1920×1080

Workflow

Manuscript → screen → showcase.

Three stages. One studio. The agent keeps every reference wired the whole way through.

  1. 01

    Write the book.

    Draft chapters in long-form Markdown with autosave. The bible suggester reads your prose. The worldmap auto-pins new named places. Your agent proposes characters, locations, lore — one-click approval before anything lands.

    Manuscript editorRevision historyDOCX / Markdown exportAuto bible suggest
    Manuscript editor screenshot1920×1080
  2. Worldmap + blueprint composite1920×1080

    Lock the world.

    Generate canonical portraits with body-view rigs. Place 3D meshes via Meshy. Build measured blueprints with key objects, light sources, camera positions. The references your shots will reuse — forever.

    Portrait rigsBlueprint editorMeshy 3DReference lineage
    02
  3. 03

    Adapt, render, publish.

    Break chapters into scenes and beats. Render the shots in the canvas. Voice the audiobook with per-character casting. Push to YouTube. Publish the world to a public showcase site — the agent keeps every bible ref wired.

    Scene editorCanvas pipelinesAudiobookPublish showcase
    Public showcase / video reel1920×1080

Pricing

Free studio. Pay for the agent. Pay-as-you-go for renders.

Three buckets. No surprises. Cancel anything anytime.

The studio
Freeforever

Everything in the app that doesn't consume API tokens. Unlimited projects.

  • Unlimited projects, chapters, scenes, bible entries
  • Manuscript editor with autosave + revisions
  • Worldmap, blueprints, screenplay, canvas
  • Audiobook script editor (manual)
  • DOCX / Markdown / Fountain export
  • Public showcase publishing
The agentRecommended
$19/ month

Unlock the in-app AI and MCP server. Fair-use tokens included.

  • Per-project agent with persistent memory
  • In-app Gemini chat across all 60+ workshop tools
  • MCP bearer token for Claude Code, Cursor, Desktop
  • Bible suggester that reads your prose
  • Reference extraction from manuscripts
  • Auto-adapter: chapter → screenplay scenes
The renders
Creditspay-as-you-go

Buy packs in-app. Spent on every render. No external API keys to manage.

$10
1,000 cr
Starter
$50
5,500 cr
+10%
$100
12,000 cr
+20%
  • Nano Banana 2 / Pro7 cr / image
  • Flux Pro10 cr / image
  • Grok Imagine9 cr / sec
  • Veo 3.142 cr / sec
  • ElevenLabs v335 cr / min
  • Meshy image-to-3D50 cr / mesh

1 credit ≈ $0.01. ComfyUI 4K upscale runs on your own GPU — no credits charged. Unused credits never expire.

FAQ

Frequently asked.

Still curious? Email hello@cinematicworkshop.com.

An AI-native studio that takes a project from blank page to finished film, audiobook, or public world. You write the novel, build the worldmap and the bible, adapt it into a screenplay, render the shots, voice the audiobook, and publish the whole universe on a dedicated showcase site — all in one place.

Every project gets its own AI agent with persistent memory of everything you've built — the bible, chapters, scenes, canvas graphs, voice casts, references, what worked and what didn't. It can draft new chapters, review what you wrote, propose bible entries from your prose, build canvas pipelines, generate and analyse images, place character voices into audiobook scripts. Same tools you have, scoped to the project it lives in.

Images: Nano Banana 2, Nano Banana Pro, Flux dev/pro, RealvisXL, plus ComfyUI upscale to 4K. Video: Veo 3.1 and Grok Imagine — image-to-video, text-to-video, first/last frame, and extend. Voice: ElevenLabs v3 with per-character casting. 3D: Meshy image-to-3D pipeline with body-view rigs (front · back · 3/4) baked into character bible entries.

Yes — each project can publish to a dedicated showcase site. Layered worldmap with click-through regions, character pages with locked portraits, location pages with blueprint floor plans, audiobook player, video reel, and YouTube upload from the canvas. The reader walks through the world you built, not just the prose.

Every named place gets a pin with real coordinates. Continents zoom into regions, regions into cities, cities into buildings, buildings into blueprints with measured floor plans. You can call workshop_measure_distance to verify travel claims across chapters — and the AI agent does this automatically as you write.

Yes. The book side is complete on its own: long-form manuscripts, chapters, scenes, the bible with characters and locations and lore, worldmap with distance verification, an editorial reviewer agent. Skip the canvas, screenplay, image, and audiobook flows if you only write prose.

Yes. Every workshop primitive is exposed as a Model Context Protocol tool — 60+ of them — so Claude Code, Claude Desktop, Cursor, and any MCP-aware client can read and write your project directly through a bearer token. Role-specific tool surfaces (write-book, write-film, write-audio, context) let you scope an external agent to one job.

Your projects live in Supabase Postgres with row-level security scoped to your user id, plus Supabase Storage for images, videos, audio, and 3D meshes. No one else on the platform can read your work, and only the AI providers you connect see the prompts you send to them.

Yes. Cinematic Workshop itself is free to use. You pay the AI providers directly for whatever images, videos, voices, and 3D meshes you generate through them, billed against your own API keys.

Yes. Manuscripts export as Markdown, DOCX, or Fountain (industry-standard screenplay format). Audiobooks export as ElevenLabs v3-ready scripts plus rendered audio. Bible entries export as JSON. Generated images, videos, and meshes are downloadable from the assets panel.

There's a mobile companion under /m for reading and lightweight editing on a phone — chapters, scenes, bible entries. The full canvas editor stays on desktop because node-graph editing doesn't work on touch.

Get in touch

Press, partnerships,
or just curious?

We read everything. Tell us what you’re building, what isn’t working, or what would make Cinematic Workshop indispensable for your next project.

Based in Stockholm. Reply within one working day.

Bring your manuscript.
We’ll set the camera.

Free to start. Subscribe for the agent. Pay-as-you-go for renders. Your work stays on Supabase, scoped to you.

Get started — free
Cinematic Workshop — From manuscript to motion picture