---
type: "post"
id: "cinecode-films-made-with-code-c0c061"
slug: "cinecode-films-made-with-code-c0c061"
title: "cinecode: Films made with code"
description: "Cinecode is an open-source workflow that asks a coding agent to turn a film idea into a rendered MP4, using code-built scenes, designed sound and review steps."
retrieval_nugget: "Cinecode is an open-source workflow that asks a coding agent to turn a film idea into a rendered MP4, using code-built scenes, designed sound and review steps. For builders, its notable choice is to make approvals and revision part of the production pipeline rather than treating video as a single prompt-and-output operation."
published_at: "2026-11-12"
updated_at: "2026-11-12"
record_date: "2026-11-12"
date_kind: "published_at"
topics: ["claude","claude-code"]
entities: ["Cinecode","Claude Code","Cursor","three.js","FFmpeg"]
related_patterns: ["generative-media-becomes-application-infrastructure"]
novelty_label: "new_fact"
what_changed: "Cinecode adds an AI filmmaking workflow that writes a screenplay, generates every frame and sound, and exports a finished MP4, extending coding-agent tools into end-to-end film production."
prior_refs: ["https://newruntime.com/posts/claude-motion-a-motion-design-toolkit-for-coding-e1ccf6/"]
source_url: "https://github.com/sarpstar/cinecode"
source_urls: ["https://github.com/sarpstar/cinecode"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:cinecode-films-made-with-code-c0c061"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/cinecode-films-made-with-code-c0c061/","markdown":"https://newruntime.com/posts/cinecode-films-made-with-code-c0c061.md","json":"https://newruntime.com/posts/cinecode-films-made-with-code-c0c061.json"}
markdown_url: "https://newruntime.com/posts/cinecode-films-made-with-code-c0c061.md"
---

# cinecode: Films made with code

## Retrieval answer

Cinecode is an open-source workflow that asks a coding agent to turn a film idea into a rendered MP4, using code-built scenes, designed sound and review steps. For builders, its notable choice is to make approvals and revision part of the production pipeline rather than treating video as a single prompt-and-output operation.

## Summary
Cinecode is an open-source workflow that asks a coding agent to turn a film idea into a rendered MP4, using code-built scenes, designed sound and review steps. For builders, its notable choice is to make approvals and revision part of the production pipeline rather than treating video as a single prompt-and-output operation.

## What happened
Cinecode’s GitHub README describes a local filmmaking tool used with Claude Code or Cursor. It is intended for short films, long YouTube videos, ads, explainers and music videos. The project says users bring the idea and creative preferences while an agent writes the script, builds frames in code, designs sound and produces an MP4. The repository labels the project MIT-licensed and identifies three.js and ffmpeg as its rendering stack. It says no paid service is required; optional external voice services include Gemini, ElevenLabs or a free Turkish voice.

The README presents two operating modes. Quick mode is aimed at work up to roughly three minutes and makes a film in a single file, with a few questions, a low-resolution draft and one or two revision rounds. Film mode is for longer, researched or multi-scene work, including documentaries and festival shorts. It can proceed through approvals at each stage or move from script to draft in one pass. The project says films can range from 15-second clips to works longer than ten minutes, with longer films requiring more time and usage.

In the described workflow, the agent asks about interpretation, medium, duration, format, visual style and sound. It then presents a beat-by-beat plan and editable director’s card; the first two scenes are shown as frames for approval before the rest is built. The draft is rendered and checked for motion, loudness, legibility and repetition, and an independent viewer is asked to review it. The README says feedback changes only the parts it concerns. The final output is full-size, mastered to −14 LUFS, with a credits file for a video description.

The repository describes several use cases beyond cinematic shorts: brand-specific product films in 9:16, 1:1 and 16:9, vertical explainers of 30–60 seconds, music videos aligned to a measured beat grid, and long-form videos with sourced claims and archival media. It also shows examples ranging from 15 seconds to five minutes. The project describes these as capabilities and examples; it does not present them as independently evaluated production results.

## What the archive adds
New Runtime’s earlier coverage of [claude-motion](https://newruntime.com/posts/claude-motion-a-motion-design-toolkit-for-coding-e1ccf6/) (October 13, 2026) described an agent-executable motion-design workflow with inspectable stages. Cinecode is a new project on the same broader idea: production work organized as a sequence an agent can execute and a person can review, rather than a one-shot generation.

## Trend signal
Cinecode supports [generative-media-becomes-application-infrastructure](https://newruntime.com/patterns/generative-media-becomes-application-infrastructure/): the README frames video as a composable production workflow, with scene code, sound, review and export, rather than just a standalone model request. This is an early project example, not evidence that the workflow is reliable or economical at scale.

## Why now
The project’s latest repository activity shown on GitHub is dated November 3, 2026. The README’s emphasis on coding-agent environments and local rendering places the tool in the current shift toward using general-purpose agents to operate creative software and code-based pipelines.

## Why it matters
For teams, the useful idea is not that an agent can replace a film crew; it is that a creative brief can become a structured, inspectable sequence of decisions and artifacts. Approval of a plan and sample scenes provides an opportunity to catch a bad interpretation before the system renders every scene. A draft review and localized revisions can also make iteration more targeted than regenerating the whole video.

That structure is especially relevant to explainers, product demos and other repeatable formats where a team can define constraints, review claims and approve visual direction. But the repository’s workflow claims are not equivalent to a demonstrated reduction in cost or turnaround time. Builders should treat Cinecode as a workflow to test against a small, bounded deliverable, with human review for factual claims, brand requirements and final quality.

## Limits and uncertainty
The repository describes features, modes and sample films, but the README does not establish independent quality benchmarks, typical completion time, compute requirements or cost per finished minute. It does not quantify how often automated checks or the independent-viewer step detect meaningful defects. Results may depend on the coding agent, local environment, chosen voice service and user approvals. The project’s descriptions of researched scripts and sourced claims do not by themselves establish fact-checking accuracy. These capabilities should be tested on representative work before teams rely on them for production.

## Skill to build now
Take one short, low-risk video brief and split it into explicit checkpoints: brief interpretation, storyboard or beat plan, sample frames, draft render and final review. Record what each checkpoint catches and how much rework it prevents. This mirrors Cinecode’s concrete approval sequence and gives a team a basis for deciding whether agent-led production improves its own workflow.

## Reader takeaway
Cinecode is best read as a workflow experiment: a coding agent can be given a staged creative-production process, not merely asked to generate a finished video. The practical test is whether its checkpoints make review and revision more manageable for a specific, repeatable format.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)