digitalsamba/claude-code-video-toolkit
13 skills12.8k installs24k starsGitHub
Install
npx skills add https://github.com/digitalsamba/claude-code-video-toolkitSkills in this repo
1FfmpegThe ffmpeg skill in digitalsamba/claude-code-video-toolkit documents FFmpeg commands for video and audio processing in Remotion production pipelines. It covers GIF to MP4 conversion with movflags faststart, yuv420p pixel format, and even-dimension scaling for web players. Resize recipes include letterbox padding, crop-to-fill, and width-based scaling. Compression presets use libx264 CRF values, AAC audio bitrates, and target file size encoding. Audio workflows extract MP3, AAC, or WAV, convert M4A for ElevenLabs samples, and adjust volume filters. Trimming guidance recommends re-encoding over stream copy to avoid keyframe seek issues. Speed change uses filter_complex with setpts and atempo, including chained atempo for extreme rates. Remotion-specific sections compare FFmpeg preprocessing versus playbackRate for constant, extreme, and variable speeds. Platform export recipes optimize masters for YouTube, Twitter, LinkedIn, web embed, and GIF previews with batch export scripts.5.6kinstalls2Playwright Recordingplaywright-recording documents the Playwright `recordVideo` browser context option from digitalsamba/claude-code-video-toolkit for automatic session video capture. Developers pass a required `dir` output path and optional `size` width/height to `browser.newContext({ recordVideo })`, alongside viewport, colorScheme, locale, timezoneId, geolocation, permissions, and userAgent settings. Reach for playwright-recording when E2E tests or agent-driven browser automation need MP4 artifacts for bug reports, demo reels, or regression forensics instead of manual screen recording. The skill is API-reference focused rather than a standalone test runner.842installs3ElevenlabsThe elevenlabs skill in digitalsamba/claude-code-video-toolkit teaches agents to generate voiceovers, sound effects, and background music through the ElevenLabs Python client with an ELEVENLABS_API_KEY in .env. It documents text_to_speech.convert with VoiceSettings presets for natural, conversational, or energetic delivery across four models: eleven_multilingual_v2 for stable 29-language output, eleven_flash_v2_5 and eleven_turbo_v2_5 for SSML breaks and phoneme tags, and expressive eleven_v3 for retake-heavy drafts. Instant voice cloning uses client.voices.ivc.create with binary sample files, while text_to_sound_effects.convert handles up to 22 second SFX and music.compose spans 10 seconds to 5 minutes with force_instrumental for beds. The toolkit ties into Remotion through tools/voiceover.py and the /generate-voiceover command, producing per-scene MP3 files plus manifest.json durations that drive Series.Sequence frame counts. Developers reach for elevenlabs when video, podcast, or game pipelines need programmatic narration, described effects, or synced soundtrack assets.817installs4Remotionremotion is a skill from digitalsamba/claude-code-video-toolkit that documents the Remotion React API for generating high-quality animated videos programmatically inside Claude Code or Cursor. Core hooks include `useCurrentFrame()` for the current 0-indexed frame—relative inside `<Sequence>`—and `useVideoConfig()` returning width, height, fps, durationInFrames, id, and defaultProps. The `interpolate()` helper maps input ranges to output ranges with extrapolation and easing options, enabling motion without manual keyframe math. Developers reach for remotion when agents need to scaffold React-based video compositions, explainer animations, or data-driven motion graphics instead of manual timeline editors. The skill is reference documentation for Remotion patterns rather than a render farm or hosting service.779installs5Qwen Editqwen-edit is a digitalsamba/claude-code-video-toolkit agent skill for AI image editing with Qwen-Image-Edit-2511 via RunPod serverless, listed on skills.sh with 519 installs. It provides prompt patterns, negative prompts, guidance and step tuning, and identity-preserving workflows for reframing cropped photos, clothing changes, pose adjustments, style transfers, and lighting edits. Developers invoke tools/image_edit.py with --prompt, --negative, --steps, and --guidance flags or multi-image composite anchors. Bundled references include prompting.md, examples.md, and parameters.md from production experiments. The skill explicitly warns against background replacement, face swapping, and outpainting where results fail reliably. Reach for qwen-edit when building media pipelines that need programmatic photo transforms with documented failure modes.655installs6Frontend Designfrontend-design is a web UI skill from digitalsamba/claude-code-video-toolkit that guides agents to build components, pages, and applications with high design quality and real working code. It explicitly avoids generic AI slop aesthetics, pushing creative typography, layout, color, and motion choices while honoring accessibility. Developers invoke frontend-design when requirements describe a component, page, or full interface needing production polish rather than wireframe placeholders. The skill expects concrete purpose and technical constraints from the user, then outputs implementable frontend code with exceptional attention to visual detail.618installs7Remotion Best Practicesremotion-best-practices is a video authoring skill from digitalsamba/claude-code-video-toolkit with 478 installs and rank 13 on skills.sh. It encodes professional Remotion patterns: animate with useCurrentFrame and interpolate instead of CSS transitions, load assets from public/ via staticFile, and embed images, video, and audio through Remotion components. The skill scaffolds new projects with npx create-video and forbids Tailwind animation classes that fail during headless renders. Developers reach for remotion-best-practices when building programmatic marketing clips, data visualizations, or agent-generated video without broken frames or non-deterministic CSS animations.618installs8Runpodrunpod is a skill from digitalsamba/claude-code-video-toolkit for running open-source AI models and video renders on RunPod serverless GPUs. It covers account setup, RUNPOD_API_KEY configuration, Docker image deployment via --setup flags, and troubleshooting endpoint issues and costs. The toolkit documents five images—qwen-edit, realesrgan, propainter, sadtalker, and qwen3-tts—for image editing, upscaling, dewatermarking, and related video tasks. Developers reach for runpod when local Remotion or MoviePy renders are too heavy and they need pay-per-second GPU endpoints with no minimum spend instead of maintaining their own GPU fleet.569installs9Ltx2ltx2 is a digitalsamba/claude-code-video-toolkit skill for AI video generation using the LTX-2.3 22B DiT model. It produces roughly 5 second clips from text prompts or source images via python3 tools/ltx2.py, running on Modal with A100-80GB GPUs and requiring MODAL_LTX2_ENDPOINT_URL in .env. Triggers include video generation, animate image, b-roll, motion content, text-to-video, and image-to-video requests inside Claude Code workflows. Developers reach for ltx2 when batching marketing reels, product demo motion, tutorial b-roll, or animated backgrounds without leaving agent-driven video pipelines.564installs10Video Toolkitvideo_toolkit is an agent skill bundled with digitalsamba/claude-code-video-toolkit for autonomous video production inside Claude Code. From a text brief it orchestrates script writing, storyboarding, AI voiceover, image generation, music, talking-head animation, and final MP4 rendering. Generation runs on cloud GPUs through Modal or RunPod using open-source models, while Remotion (React) handles timeline composition and export. The skill requires node, python3, ffmpeg, and npm on darwin or linux hosts and expects agents to cd into the fixed toolkit path before invoking pipeline commands. Developers reach for video_toolkit when creating product demo reels, tutorial shorts, launch clips, or narrated explainers without leaving the repository. It pairs with the toolkit /setup and /video commands for GPU configuration and first-render workflows.555installs11Moviepymoviepy is a skill in digitalsamba/claude-code-video-toolkit for Python video composition with MoviePy 2.x. It targets the gap where LTX-2 and SadTalker cannot render trustworthy on-screen text—names, prices, and attributions must be pixel-accurate—by compositing PIL-rendered ImageClips over AI footage. The toolkit ships two runnable examples, quick-spot and data-viz-chart, plus patterns for lower thirds, tinted contrast overlays, side-by-side composites, and audio-anchored VO timelines under 30 seconds. Developers reach for moviepy when captioning SadTalker talking heads, labeling LTX-2 b-roll, or building single-file build.py ad spots without a Remotion or Node.js toolchain, reserving Remotion for long-form React-driven productions.550installs12Acestepacestep is a digitalsamba Claude Code video toolkit skill for AI music generation with ACE-Step 1.5. Open-source generation runs through tools/music_gen.py with cloud providers including acemusic as the default official API using the XL Turbo 4B model plus 5Hz LM thinking mode. Developers reach for acestep for background music, vocal tracks, covers, stem extraction, audio repainting, continuation, jingles, and style transfer inside video production workflows. Free API keys come from acemusic.ai and triggers include soundtrack, music generation, stem extraction, cover, repaint, and composition tasks.518installs13Ideogram4Generates images with legible on-image text and precise color/layout control via Ideogram 4's structured JSON caption format. A developer uses it to make title cards, thumbnails, and signage with baked-in text.164installs