One download.
Room to make it yours.
An organized local video toolkit with both Codex and Claude Code adapters. The same source and tools make the examples you see here. All nine styles and the local creative draft and thumbnail helper are included at $67 once.
Try the presenters and stylesWhat you receive
| Folder | What’s inside |
|---|---|
| Nine styles | Soft 3D, Detailed Pixel Art, Floating Cards, Cut-Paper Stories, Comic Panels, Blueprint Explainers, Plushy, Cinematic Storybook and Documentary. Native horizontal and vertical example films, editable source, supplied speech, captions and sound. Explore all nine. |
| Start here | A narrated first-video walkthrough, copyable commands, both assistant adapters and setup instructions. Follow the first-video guide. |
| Starter projects | Three complete originals: an illustrated explainer, a fictional product demo and an animated practice-recording edit. Each has both native formats, editable source, narration, timings, captions, sound and rights records. |
| Brand & delivery | Reusable supported brand settings, an export checklist and a local helper that collects movies, captions, covers, source and declared rights with file hashes. |
| Studio | Three finished designs, eleven presenters, six backgrounds, four fonts and a local browser preview with MP4 rendering. |
| Motion & story | 100 motion effects, 12 Story layouts in three looks, 16 kinetic worlds and 22 Craft worlds. Layout/look combinations are options, not extra styles. The audio waveform is illustrative, not soundtrack analysis. |
| Recordings | 35 recording treatments, overlays and an optional local macOS segmentation helper. |
| Voice & sound | Local narration and word alignment tools, adjustable speed and pauses, volume normalization, a lighter voice treatment and 3 reusable stock voice presets, 12 original procedural accents plus 53 credited CC0 recordings across 15 categories, with frame-aligned peak placement. |
| Examples & source | The new presentation and full walkthrough, the real rainbow before/after, both formats and editable project files. |
| Creative drafts & thumbnails | Local rule-based idea, hook, causal storyline and shot-plan starters. Three actual 1280×720 PNG thumbnails, editable SVG and wording. Optional assistant-authored writing uses the same schema. See the actual generated files. |
| Documentation | Workflow guides, licenses and notices. Earlier examples remain available in a clearly marked legacy section. |
Future updates to this video toolkit are included. Optional services, separate future products and new third-party assets are not automatically part of the purchase.
What setup involves
- Install Node.js 20.9+, FFmpeg/FFprobe with libass captions and Python 3.11+ (3.12 tested), then install the toolkit’s pinned rendering dependencies. The three sentence-led style finishers need Python 3.11+, NumPy and Pillow; their exact pins are supplied separately. Chrome or Chromium is used by the style renderer.
- For optional local planning, narration and alignment, install Python packages and download the pinned public models. The models need roughly 6.7 GB, plus extra space for setup and projects.
- Open the included local Studio. Choose a presenter, background and font, preview the example, then render an MP4. Use a local assistant or the documented commands to build a new project.
The full local model workflow has been exercised on an Apple Silicon M1 Pro Mac with 32 GB memory. Rendering and fresh-package checks were performed on that machine. Other operating systems, lower-memory hardware and an independent buyer installation remain unverified.
Costs, plainly.
The toolkit price is $67 once. After purchase and setup, rendering each video on your computer costs zero dollars. You can save the rendered video files on your own computer. AI help and premium media can cost extra. Your computer, electricity, storage and initial downloads are your costs. ChatGPT/Codex or Claude usage draws on your own plan, limits or credits and is billed separately by that provider. Cloud narration, stock media and generated footage are optional and separate.
The new presentation uses local narration and original vector graphics. Its reusable customer voice preset is original. The finished example audio is included; the private reference recording and speaker profile are excluded. Use permission-cleared references if you make a voice clone.
Ask for the change.
Once setup is complete and your local assistant has loaded the included skill, use plain language: “Change the script,” “Use Eva,” “Make the captions bigger,” or “Use a paper background.” Your assistant edits the project files; preview, review and render the result. The templates and presets come organized in the download. Node.js 20.9+, FFmpeg/FFprobe with libass captions, Python 3.11+, Chrome or Chromium and project dependencies need initial setup. Assistant access and optional local models are separate.
Caption choices
Use progressive word-timed captions, coloured keyword emphasis or the recording caption/card treatments. Standard English captions can be switched on in the player, and SRT/VTT files come with each export. Ask the assistant to change the font, size, colour and timing, then check the chosen format for readable framing.
Your characters. Your recordings.
Create or customize a vector presenter in the editable source with your coding assistant, starting from eleven included presenters. Bring your own local videos into the recording treatments and overlays. There is no automatic photo-to-avatar builder or cloud upload service.
Which AI runs it?
Rendering supplied scripts and audio needs no AI model. Optional local planning uses mlx-community/Qwen3-4B-Instruct-2507-4bit; narration uses mlx-community/Qwen3-TTS-12Hz-1.7B-Base-8bit; word timing uses mlx-community/Qwen3-ForcedAligner-0.6B-8bit. Pinned downloads and commands are included. These models are separate initial downloads, about 6.7 GB in total.
Alternatively, use a local Codex or Claude Code assistant with file access and the included adapter. You provide that assistant access. Ordinary web-chat file uploads alone are not the tested setup.
What the examples demonstrate
Word-timed 2D motion, articulated vector presenters, source editing, local voice and local MP4 exports. The Studio vector presenter’s mouth uses sound level and approximate word shapes; this is not phoneme-accurate lip sync. The three sentence-led style sources are finite authored scenes. New subjects and stories need source edits; arbitrary uploaded characters are not automatically rigged. Some sample recording footage is synthetic and labeled as such.
Detailed rendering and model instructions are in the download. The full effects gallery and existing thumbnail tools stay available here as supporting resources.
Open the full gallery · Open thumbnail tools · Watch the walkthrough
Current toolkit: 0.13.0
The current toolkit includes both local assistant adapters, all nine styles and the one-prompt setup guide. The product is available on Gumroad for $67 one-time. Review the setup requirements and current product policies before purchase. Open the setup prompt · Read the policies.