digitalsamba/claude-code-video-toolkit: AI-Native Video Production in Claude Code

digitalsamba/claude-code-video-toolkit: AI-Native Video Production in Claude Code

Creating professional video content typically means leaving your development environment. You switch to separate tools for scripting, voiceover, asset generation, and editing — each with its own subscription, learning curve, and export-import friction. For developers who live in code and want to automate explainer videos, sprint reviews, or product demos, this context-switching tax is especially painful.

The digitalsamba/claude-code-video-toolkit solves this by bringing the entire video production pipeline into Claude Code. You describe what you want in natural language, and Claude Code handles scriptwriting, voiceover generation, music, visuals, and final MP4 rendering — all within your familiar terminal-based workflow. This claude-code-video-toolkit is built around open-source AI models that cost cents per video, not dollars, and it's designed specifically for technically literate users who want reproducible, version-controlled video projects.

What is digitalsamba/claude-code-video-toolkit?

The digitalsamba/claude-code-video-toolkit is an AI-native video production toolkit maintained by Digital Samba, a company focused on AI-powered mobile development. The project lives at https://github.com/digitalsamba/claude-code-video-toolkit and carries an MIT License. With 1,746 GitHub stars, 296 forks, and Python^{↗ Bright Coding Blog} as its primary language, it has gained meaningful traction in the developer-tools community since its initial release.

The toolkit's last commit was on 2026-07-06, indicating active maintenance. Its core premise is treating Claude Code not just as a coding assistant, but as a video production agent — equipped with domain-specific skills, slash commands, templates, and Python-based tools for media generation. The author describes it as "a reusable toolkit for using Claude Code to autonomously generate any kind of 'explainer' style video."

What makes this approach distinctive is its agent-native architecture. Rather than wrapping a GUI video editor in API calls, the toolkit gives Claude Code structured capabilities: skills for Remotion (React^{↗ Bright Coding Blog}-based video composition), ElevenLabs and Qwen3-TTS for voice, FLUX.2 and Ideogram 4 for images, ACE-Step for music, and LTX-2 for AI video clips. Claude Code becomes the director; you become the editor and designer who steers intent.

The cost structure is deliberately developer-friendly. Open-source models run on your own cloud GPU account — voiceovers cost approximately $0.01, AI video clips around $0.23. Cloudflare R2 offers 10GB free storage with zero egress fees, and Modal's Starter plan provides $30/month in free compute, which the documentation notes is sufficient for several 5-minute videos monthly.

Key Features

The toolkit organizes its capabilities into skills, commands, templates, and Python tools — a modular system that lets you invoke exactly what you need without loading unused dependencies.

Skills extend Claude Code's domain knowledge across ten areas: Remotion for React-based video composition; ElevenLabs for AI audio; FFmpeg for media processing; Playwright for browser demo recording; frontend-design for visual refinement; Qwen-Image-Edit and Ideogram 4 for image editing and generation with best-in-class in-image text; ACE-Step for music generation; LTX-2 for AI video generation; MoviePy for Python-based video composition; and RunPod for cloud GPU management.

Slash commands provide the interactive interface. /setup walks through cloud GPU, storage, and voice configuration interactively. /video creates or resumes projects. /scene-review opens Remotion Studio for frame-by-frame inspection. /design triggers focused visual refinement sessions. /generate-voiceover, /voice-clone, and /redub handle audio workflows. /publish uploads finished videos to YouTube with metadata auto-populated from project.json. Other commands manage brand profiles (/brand), templates (/template), skills (/skills), and demo recording (/record-demo).

Templates provide pre-built video structures: sprint-review and sprint-review-v2 for development team updates; product-demo for marketing with dark tech aesthetics; and concept-explainer-short for 9:16 vertical shorts using Python/MoviePy without Remotion.

The project management system tracks video projects through a multi-session lifecycle: planning → assets → review → audio → editing → rendering → complete. Each project maintains a project.json that reconciles planned intent against actual filesystem state, plus a CLAUDE.md for instant context when resuming work across sessions.

Scene transitions include seven custom effects — glitch, rgbSplit, zoomBlur, lightLeak, clockWipe, pixelate, and checkerboard — plus four official Remotion transitions. These are previewable through a dedicated showcase project.

Use Cases

Sprint review videos for distributed teams. The toolkit's original use case, documented in the author's note: creating sprint review videos for Digital Samba's AI mobile development team. The sprint-review and sprint-review-v2 templates structure demos, statistics, and voiceover into polished updates. Playwright recording captures browser-based features; Claude Code assembles and narrates.

Product explainer shorts for technical products. The concept-explainer-short template generates 9:16 vertical videos optimized for TikTok, Reels, and Shorts. The documented "Why is the sky blue?" example runs 52 seconds and costs approximately $0.80 in generation fees — demonstrating viability for high-volume content pipelines.

Marketing launch videos with cinematic production value. The "Super Bowl-style launch ad" example combines LTX-2 animated cameos, dramatic Qwen3-TTS announcer voice, and structured scene composition. The product-demo template provides dark tech aesthetics with stats and call-to-action elements.

AI-generated video essays and educational content. "The Space Between" example shows a flux2 avatar with Qwen3-TTS voice and SadTalker lip-sync animation — suitable for long-form YouTube content or course materials.

Automated demo recordings with branded post-processing. The /record-demo command captures browser interactions via Playwright, while notebooklm_brand.py and related utilities rebrand existing content with custom logos and URLs.

Installation & Setup

The toolkit requires Node.js 18+, Claude Code, and Python 3.9+ recommended for AI tools. FFmpeg is optional.

Basic installation:

git clone https://github.com/digitalsamba/claude-code-video-toolkit.git
cd claude-code-video-toolkit
python3 -m pip install -r tools/requirements.txt   # Optional: AI voiceover, image gen, music, moviepy examples
claude                                              # Open Claude Code in the toolkit

The git clone pulls the full repository including skills, commands, templates, and tools. The pip install step is marked optional because core Remotion-based rendering works without Python dependencies — they're only needed for AI voiceover, image generation, music, and MoviePy examples.

First-time configuration inside Claude Code:

/setup                    # Configure cloud GPU, storage, voice (~5 min, mostly free)
/video                    # Create your first video

/setup is interactive, walking through cloud GPU provider selection (Modal recommended, RunPod supported), file transfer configuration, and voice setup. The "mostly free" characterization reflects Modal's $30/month Starter plan credit and Cloudflare R2's 10GB free tier.

Immediate render without setup:

cd examples/hello-world && npm install && npm run render

This produces an MP4 with no API keys required — useful for verifying the toolchain before committing to cloud GPU configuration.

For cloud GPU deployment, run /setup to automatically deploy all tools to Modal, or use python3 tools/<tool>.py --setup for RunPod endpoint creation.

Real Code Examples

The toolkit's Python CLI tools follow consistent patterns. Here are documented examples from the README:

AI voiceover with Qwen3-TTS (self-hosted, lower cost):

# Generate voiceover with specific speaker and structured output
python tools/voiceover.py --provider qwen3 --speaker Ryan --scene-dir public/audio/scenes --json

This uses Qwen3-TTS instead of ElevenLabs for cost reduction. The --json flag outputs structured metadata alongside audio files. The --scene-dir parameter organizes outputs per-scene for multi-scene video projects.

AI music generation with ACE-Step (free cloud API):

# Generate 120 seconds of corporate background music
python tools/music_gen.py --preset corporate-bg --duration 120 --output music.mp3

# More specific musical direction with key and BPM
python tools/music_gen.py --prompt "Dramatic cinematic" --duration 30 --bpm 90 --key "D Minor" --output reveal.mp3

ACE-Step offers a free cloud API tier, making this genuinely zero-cost for basic usage. The preset system covers common video scenarios; custom prompts allow precise control.

AI image generation with FLUX.2 and brand integration:

# Generate title background using brand colors and assets
python tools/flux2.py --preset title-bg --brand digital-samba --cloud modal

# List available presets to discover options
python tools/flux2.py --list-presets

The --brand parameter pulls from brands/digital-samba/brand.json for consistent colors and typography. The --cloud modal flag routes to Modal's serverless GPU infrastructure.

AI video generation with LTX-2.3:

# Text-to-video generation
python tools/ltx2.py --prompt "A sunset over the ocean, cinematic" --cloud modal

# Image-to-video with camera motion
python tools/ltx2.py --prompt "Gentle camera drift" --input photo.jpg --cloud modal

LTX-2.3's 22B parameter model handles both text-to-video and image-to-video workflows. The estimated cost of ~$0.23 per clip makes it feasible to generate multiple b-roll options and select the best.

Advanced Usage & Best Practices

Based on the toolkit's design patterns, several practices emerge for effective use:

Start with templates, then customize. The examples/ directory contains 10+ finished projects with documented costs and techniques. Study these before building from scratch — the "hallucinations-short" example demonstrates how Ideogram 4 cards, LTX-2 b-roll, and karaoke-style captions combine for educational content.

Use brand profiles for consistency. The brands/ directory structure (brand.json, voice.json, assets/) ensures repeatable visual identity across projects. Create your brand once with /brand, then apply automatically to all subsequent videos.

Leverage the session resumption system. The multi-session lifecycle with CLAUDE.md context files means you can start a project, pause for hours or days, and resume with full context. Don't try to complete complex videos in single sessions.

Prefer Modal for predictable costs. While RunPod offers pay-per-second flexibility, Modal's $30/month credit and automatic container management reduce operational overhead. Typical usage runs $1-2/month according to documentation.

Validate with /scene-review before rendering. Remotion Studio provides frame-accurate preview; catching timing issues here avoids expensive re-renders of final MP4 output.

Comparison with Alternatives

The toolkit occupies a specific niche: developers who already use Claude Code and want to stay in that environment while automating video production. Descript offers lower setup friction but less automation potential. Raw Remotion requires building your own AI integrations. Runway provides superior individual clip quality but no agent-based workflow orchestration.

FAQ

Does this require paid API keys? No. The toolkit leans heavily on open-source models and free tiers. Modal's $30/month Starter credit, Cloudflare R2's 10GB free tier, and ACE-Step's free API make basic usage genuinely zero-cost.

Can I use this without Claude Code? Partially. The Python tools work standalone, and an experimental Codex migration script exists (scripts/migrate_to_codex.py). However, the full agent-native workflow requires Claude Code's skill and command system.

What video quality can I expect? Depends on source assets. LTX-2.3 generates 720p clips; final output resolution is controlled in Remotion composition settings. The examples demonstrate professional results with proper asset curation.

Is voice cloning included? Yes. /voice-clone records, tests, and saves cloned voices to brand profiles. Qwen3-TTS supports cloning from reference samples; ElevenLabs integration is also available.

How do I publish to YouTube? Use /publish inside Claude Code or python tools/youtube_upload.py directly. OAuth 2.0 setup is one-time; metadata auto-populates from project.json.

What's the learning curve? Moderate. If you know React, Remotion's composition model is familiar. The /setup and /video commands guide beginners. The examples/hello-world renders without any configuration.

Can I contribute? Yes. The author explicitly welcomes issues and PRs. See CONTRIBUTING.md for guidelines.

Conclusion

The digitalsamba/claude-code-video-toolkit is purpose-built for a specific persona: developers and technical teams who want to automate video production without leaving their primary workflow. It sacrifices the immediate accessibility of GUI tools for the reproducibility, version control, and cost efficiency of code-native pipelines.

With 1,746 stars and active maintenance through mid-2026, it has demonstrated staying power. The open-source model approach — Qwen3-TTS, FLUX.2, ACE-Step, LTX-2 — keeps per-video costs in cents rather than dollars. The agent-native architecture, where Claude Code orchestrates rather than merely assists, represents a genuine paradigm shift for technical content creation.

If you're already using Claude Code, have basic React familiarity, and produce regular explainer videos, sprint reviews, or product demos, this toolkit merits serious evaluation. Start with the examples/hello-world instant render, then progress through /setup to unlock the full AI generation pipeline.

Explore the repository at https://github.com/digitalsamba/claude-code-video-toolkit and see what Claude Code can direct for you.