# Shotcoda

> Shotcoda is an agentic AI video pipeline that turns an idea into a finished, uploaded YouTube Short (1080×1920, 30 fps, −14 LUFS). It researches the topic, writes the script, designs a narrator voice, generates and animates every shot, composes music, edits with captions and motion graphics, runs a quality gate and uploads to YouTube, Facebook and Instagram. All media is generated locally on the owner's NVIDIA GPU with open-weight models; no video SaaS, stock footage or per-video fees. Built by Harri Reili; a product of Mervantis OÜ (Estonia). Contact: hello@shotcoda.com.

Key facts (measured on the reference machine, 2 × RTX 5060 Ti 16 GB, Ryzen 5 7600X, 32 GB RAM, Ubuntu):
- Fast profile: 48 minutes of machine time for a 62.6-second Short (12 AI clips, 4 stills, 4 motion graphics).
- Production profile: about 1 h 45 min per 60-second Short (21 AI-animated shots).
- Minimum hardware: one NVIDIA GPU with 16 GB VRAM (RTX 4060 Ti 16 GB / 5060 Ti 16 GB / 4070 Ti Super), 64 GB RAM recommended, 150 GB SSD, Linux. Motion takes roughly twice as long on one GPU. GPUs under 12 GB VRAM, AMD GPUs and macOS are not supported.
- Models: Qwen3.8-27B (director + QC judge, or Claude/ChatGPT), Qwen3-TTS 1.7B (voice), faster-whisper (timing), Z-Image Turbo (keyframes), Wan 2.2 I2V A14B + lightx2v (motion, split across two GPUs), ACE-Step 1.5 (music), RIFE v4.6 (interpolation), HyperFrames + GSAP (edit), FFmpeg (mastering), ComfyUI (model runtime). About 89 GB of model files.
- Shotcoda is not a SaaS or sign-up app: it is built and installed on the customer's own hardware. No public pricing; enquiries via the contact page.

## Pages
- [Home](https://shotcoda.com/): what Shotcoda does, example Shorts, hardware summary, FAQ
- [How it works](https://shotcoda.com/how-it-works/): all twelve stages with engines, devices, timings and outputs; quality control
- [Showcase](https://shotcoda.com/showcase/): published Shorts with machine time, shot counts and QC results
- [Hardware](https://shotcoda.com/hardware/): GPU/RAM/storage tiers and measured per-stage VRAM and timings
- [Start here](https://shotcoda.com/start/): free starter guide, brief template, hardware check
- [About](https://shotcoda.com/about/): founder Harri Reili and Mervantis OÜ
- [Contact](https://shotcoda.com/contact/): enquiry form and email

## Example videos
- [Saturn Hexagon Has a New Twin](https://www.youtube.com/shorts/KruAmfmzK8o): fast profile, 48 min machine time
- [The 52 Hertz Whale: Heard for Decades, Never Seen](https://www.youtube.com/shorts/KCM5wypLwGU): first production run

## Channels
- YouTube: https://www.youtube.com/@Shotcoda
- Facebook: https://www.facebook.com/Shotcoda/
- TikTok: https://www.tiktok.com/@shotcoda
- Founder: https://hreili.com · https://www.linkedin.com/in/harri-reili-3891068a/