IDOCInTheBox
v6.1
IDOCInTheBox
v6.1
The complete manual — product guide, capture guide, AI feature guides, plugin development reference and CLI reference — ships inside the application under Help › Documentation. This page covers what you need before and just after install.
The shortest path from a tape to a restored file.
Keep your originals. Always retain the original capture and the physical tape. Restoration is judgment work, and you will want to revisit decisions later with better settings.
| Operating system | Windows 10 or Windows 11, 64-bit |
|---|---|
| Memory | 8 GB minimum • 32 GB recommended for 4K work |
| GPU | NVIDIA with CUDA strongly recommended for the AI engines. DirectML provides a vendor-neutral fallback on other GPUs and NPUs; CPU inference works but is slow. |
| VRAM | 2 GB minimum • 8 GB or more for 4K and the heavier engines |
| Disk | 10 GB for the application and AI models, plus working space for footage — budget generously, intermediate files are large |
| Capture hardware | Any DirectShow-compatible capture device for analog, or a FireWire/DV interface for DV |
Device selection, crossbar input (Composite, S-Video, Component), codec choice across seven archival formats, live monitoring during capture, and DV deck transport control. In-app: Help › Capture Guide.
Deinterlacing, denoising, deblurring, upscaling, frame interpolation, colorization and face restoration — what each engine is for, when it helps, and when it hurts. In-app: Help › AI Enhancement Guide.
Reading VMAF, PSNR and SSIM, and using preset comparison to choose between candidate settings on evidence rather than impression.
LavaSR v2 speech restoration and Echoes Over Time enhancement — recovering intelligible dialogue from degraded analog audio tracks.
Writing community plugins against the three pipeline hooks, the settings schema, and the per-frame filter API. In-app: Help › Plugin Development Guide.
Batch and headless operation for larger archival projects, including scripted preset application across many tapes.
Some parts of the project are public:
Open the Diagnostics page. Every engine reports its own status — Ready, Missing Dependencies, or Import Error — with the specific dependency named. Use Copy Support Report and include the output if you contact support.
The device scheduler assigns engines across CUDA, DirectML and CPU based on available VRAM, but stacking several heavy engines at 4K can still exhaust it. Reduce output resolution, run fewer engines per pass, or let the scheduler move an engine to CPU.
Most capture failures are the device rejecting the requested format rather than a driver problem. Some USB capture dongles accept only exactly 30 or 25 fps and will refuse 29.97 drop-frame. Try the nearest whole-number frame rate first.
Email [email protected]. Include your order ID, your Windows version, your GPU, and the support report from the Diagnostics page — that combination resolves most issues on the first reply.