What local rendering means in practice
When TubeForge renders a video, it runs FFmpeg directly on your machine. The render pipeline generates a filtergraph (scene clips, motion effects, audio mix, captions, transitions) and executes it through the bundled static FFmpeg binary — no internet connection required, no render queue, and no cost per export. TubeForge's Adaptive Rendering Engine probes your hardware at first run, picks the fastest verified encoder (AV1 on modern hardware, H.265 or H.264 as fallback), validates it with a one-frame test encode, and caches the result. If a GPU encoder misbehaves mid-render, TubeForge automatically retries with CPU libx264 so the render completes even on tricky hardware. Re-rendering to tweak a motion effect or caption style costs nothing but a few minutes of electricity.
Privacy: what leaves your machine, and what doesn't
TubeForge is designed to minimise what leaves your machine. The only things that egress from the desktop app are: AI generation prompts sent to your provider on your plan credits (OpenRouter), the finished video you explicitly choose to upload to YouTube via your own OAuth, and optionally your channel metadata to your own Supabase cloud-sync instance if you enable it. Source media (images, voices, renders), your scripts, your brand strategy, and your unreleased content stay on your disk. The app sends no telemetry, no crash reports, and no usage analytics from the desktop. The renderer network is locked to `127.0.0.1` (loopback only) in production builds.
Offline animated captions
Animated captions are aligned using the bundled faster-whisper model, which runs entirely offline on your machine — no audio upload to a transcription API. This is significant for local-first use: you can generate word-level captions on unreleased content without sending audio to any external service.
Local-first capabilities
- FFmpeg render on your own GPU — no shared cloud render farm, no queue.
- Unlimited re-renders at no cost — only electricity, not a per-export meter.
- No watermark ever — local render means no watermark gate.
- Project database is a local SQLite file on your disk — no account needed to produce a video.
- Secrets in the OS keychain — API keys never stored in the app database or logs.
- Animated captions aligned offline — no audio leaves the machine for transcription.
- No telemetry, no crash reporting from the desktop.
What a local AI video editor trades off
Honest trade-offs: TubeForge is a desktop app, so you need to install it and add your AI keys before making your first video (roughly 10 minutes of setup). It is not a general non-linear editor — it is a production pipeline for narration-driven faceless content (script → scenes → voice → visuals → render → upload). If you want to edit arbitrary footage on a timeline, you need a different tool. And if you don't have a GPU at all, render falls back to CPU libx264, which is slower (minutes per scene rather than seconds) but still works and still costs nothing per export. The pipeline is optimised for faceless YouTube channels, not for vlog editing, live-stream clipping, or short-film production.
Try TubeForge on your own machine
TubeForge is a local-first desktop app for Windows 10/11 and macOS 11+ (Apple Silicon & Intel). Bring no API keys, render on your own GPU with bundled FFmpeg, set a per-video spend cap with Profit Mode, and keep every project on your disk. Grab the installer below.
Free tier + plans from $9/mo · no API keys · install guide
Frequently asked questions
- Does TubeForge require an internet connection to edit and render?
- Only for AI generation (scripting, images, voice, video clips — which need your AI provider key) and for YouTube upload. The FFmpeg render itself is entirely local and offline — you can re-render an existing project with no internet connection, which is useful for testing motion tweaks or caption styles without incurring AI costs.
- What hardware does TubeForge need for local rendering?
- Windows 10/11 or macOS 11+ (Apple Silicon and Intel both supported). A discrete GPU (NVIDIA, AMD, Intel Arc, or Apple Silicon GPU) is recommended for fast renders — the Adaptive Rendering Engine picks the fastest verified encoder on your hardware. Without a discrete GPU, TubeForge falls back to optimised CPU rendering (libx264) which works on any modern machine, just slower.
- Is TubeForge a general video editor?
- No. TubeForge is a production pipeline for narration-driven faceless YouTube content — idea → script → voice → visuals → motion → render → upload. It is not a timeline-based non-linear editor for arbitrary footage. If you need to cut your own footage or do live-stream clipping, a tool like DaVinci Resolve or CapCut is the right fit for that work.
- Can I use TubeForge without uploading my scripts to any cloud?
- Yes. By default, TubeForge is entirely local: your script, storyboard, brand bible, and unreleased content stay on your machine's disk. The only things that leave the machine are AI prompts sent to your provider (under your key, not proxied) and the finished video you choose to publish. Cloud sync is an optional, opt-in feature — you can use the full studio without it.