IDOCInTheBox
v6.2
IDOCInTheBox
v6.2
Advanced Tape Restorer v6.2 is the most powerful Windows application for rescuing VHS, Hi8, 8mm, Betamax, and DV tapes — with a 21-engine AI suite covering upscaling, interpolation, face restoration, colorization, audio enhancement, damage repair, and more.
Instant license key delivery by email. 30-day money-back guarantee.
Real family tapes and film scans — captured, restored, and upscaled with ATR's AI pipeline. Nothing staged.
Every tape runs through a multi-stage professional pipeline — each engine layering on the last for maximum quality.
Version 6.2 adds an in-house face restoration engine you can personalize to your own family, Blackmagic RAW film-scan ingest, and a color-managed path all the way into DaVinci Resolve.
A face-restoration engine built and trained entirely in-house — no third-party model licenses attached to your output. Three variants ship with the app: v2 for general video, v2r tuned for analog tape, and v2f for scanned film — and the pipeline routes each source to the right one automatically.
New — Face AITrain ATR Face Restore on the faces that matter most — your own family. The Reference Library's built-in trainer fine-tunes a personalized model from your reference photos, ships in three editions matched to your GPU, and the pipeline auto-detects and uses your model whenever it's present.
New — TrainingRestore scanned film, not just tape. Select a .braw file from a film scanner and a BRAW develop card appears — gamma and gamut controls up to Rec.2020 plus orientation fixes for flipped or upside-down scans, with full-quality BRAW decoding bundled in the app.
A new Color Management card adds a BT.2020 wide-gamut working space, and HDR (PQ / HLG) sources are tone-mapped instead of clipped. Color is tracked through the entire pipeline, and every output carries correct colorspace tags.
New — ColorFinish in Resolve without guesswork. The Resolve Wide Gamut Master preset writes a DaVinci Wide Gamut / Intermediate mezzanine, and the color-managed handoff opens your restoration in Resolve with matching project settings applied.
New — NLEThe Reference Library now runs on a commercial-safe detection stack (YuNet + SFace), clusters detected faces so you name each person once, and assigns identity per face-track with an ambiguity margin — fewer wrong matches on blurry tape footage.
New — FacesPlus: a reworked live capture monitor (verified on real hardware), an upgraded in-app player, a seeded troubleshooting knowledge base, and the full atr-cli in the packaged app. Full release notes →
From SD-to-4K upscaling and frame interpolation to B&W colorization, face restoration, audio enhancement, and tape artifact repair — every restoration challenge has a dedicated engine.
ATR automatically picks the fastest available backend — or you can override per-engine.
Built on VapourSynth — the professional video processing framework. Every filter is tunable from the GUI, with live preview so you can see results before committing a full render.
Tape audio degrades along with video — muffled frequencies, hiss, and degraded speech. ATR v6.2 includes two dedicated AI audio engines plus a chain mode that combines both.
Upsamples degraded 16kHz tape audio to full 48kHz broadband using a HuggingFace AI model. Includes a built-in denoising toggle to remove tape hiss before expansion. Runs on CUDA, CPU, or DirectML.
Audio AIVoiceFixer-based speech restoration with a dry/wet mix slider (0.0–1.0) so you control how aggressively the AI reshapes the audio. Ideal for interviews, family conversations, and narration on old tapes.
Audio AILavaSR cleans up broadband quality first, then Echoes Over Time refines speech clarity in a second pass. One-click selection in the engine dropdown. Automatic audio detection skips enhancement when no audio stream is found.
Audio AIResemblyzer-based 256-d d-vector speaker embedding lets you identify and match speakers across segments of your restored audio. Gracefully skips when Resemblyzer is not installed.
ATR v6.2 goes beyond traditional restoration with three face-restoration engines — including our own trainable ATR Face Restore — plus automatic colorization of black & white footage.
New in v6.2: our in-house face engine restores faces without inventing detail that was never there — and it's the only one you can train on your own family. Three shipped variants (v2 video, v2r tape, v2f film) are routed automatically by source type, and the in-app trainer builds a personalized model from your reference photos on your own GPU. Trained entirely in-house, so no third-party model license touches your output.
New in 6.2 Face AIBuilt-in GFPGAN v1.4 (350 MB, Apache-2.0) detects and enhances faces automatically. CodeFormer is available as a user-installed alternative (S-Lab non-commercial license¹) with an adjustable fidelity slider for fine-tuned control between realism and sharpness. Both support adjustable detection threshold and separate background enhancement.
Face AIBuild a personal face identity library from your own photos — imported from local files, iCloud® Photos, Google Photos™ (Takeout), or Amazon Photos desktop. ATR auto-detects your iCloud sync folder, clusters faces with scikit-learn, and matches against tape footage using an adjustable similarity threshold (0.40–0.80). All processing is fully local — photos are never uploaded.
Face ReferenceAutomatically detects black & white footage and colorizes it using DeOldify's 120 MB AI model. An artistic factor slider (0–35) controls colorization style from conservative to vivid. Choose between Stable video and Artistic models.
Colorization AIUses temporal context from 10+ surrounding frames to reconstruct exactly what should be in a damaged area — removing scratches, mold spots, dropouts, tracking errors, date stamps, and watermarks. Three modes: Remove Artifacts (auto), Object Removal, and Restore Damaged Areas. (Requires separate installation — Apache-2.0 license.)
Inpainting AIATR v6.2 includes professional-grade video quality metrics — the same standards used by Netflix, broadcasters, and streaming platforms — with per-frame CSV and PNG export.
Every metric is computed per-frame and exported to CSV and PNG graphs — so you can pinpoint exactly which frames in your tape had the worst damage and verify your restoration improved them.
AnalysisAutomatic shot boundary detection with adjustable threshold and minimum scene length filters. Supports multiple detection algorithms to handle tape sources with inconsistent cuts.
AnalysisGenerate detailed HTML quality reports covering resolution, framerate, codec, interlace pattern, field order, scene list, and all quality metrics — shareable with clients or archivists.
AnalysisATR v6.2 includes a full hardware capture module — no third-party capture software needed. Auto-detect your capture card, digitize your tape losslessly, then restore — all without leaving the app.
DirectShow-based capture for USB and PCI analog capture cards. Supports Composite, S-Video, and Component input. Configurable for NTSC (720×480 @ 29.97fps) or PAL (720×576 @ 25fps). Lossless codecs: HuffYUV, Lagarith, FFV1, UT Video.
Built-inCaptures DV and miniDV tapes via FireWire (IEEE 1394) with native stream copy mode — preserving the original DV encoding with minimal loss. Supports miniDV, HDV, DVCAM, and DVCPRO. Timecode extraction included.
Built-inAutomatically enumerates all connected DirectShow devices, classifies them as analog or DV, and displays them in a dropdown — no manual driver configuration required. Supports multi-device setups.
CaptureATR v6.2 is a full restoration suite — not just a processor. The built-in scene editor, timeline tools, and batch scheduler handle your entire archive workflow.
Real-time split-screen comparison with an adjustable divider between original and processed. Frame-by-frame navigation, zoom control, scene creation/merging/splitting, and synchronized timeline playback. Quickly verify every frame before final export.
Scene EditorSet precise start and end points with I/O keyboard shortcuts or draggable handles. Process only a portion of your tape — with a right-click context menu, focus-aware key handling, minimum gap enforcement, and exact audio/video trim sync via FFmpeg -ss/-to.
Queue unlimited jobs, apply presets to the entire batch, and monitor per-job ETA in real time. Schedule a delayed batch start for overnight processing. SMTP email notifications on completion and individual job errors — with one-click test from Settings.
BatchThree pipeline hook points — pre_filter, post_filter, and post_encode — let community plugins inject custom per-frame processing. Plugins auto-discovered from %LOCALAPPDATA%, with dynamically generated settings UI, drag-to-reorder, cancel support, and skip-on-failure safety.
One click to apply a full restoration configuration. Save your own and share with others.
Search royalty-free music via Jamendo (Creative Commons), generate custom music from text prompts via ElevenLabs AI, or run fully local AI music generation with ACE-Step 1.5. Download to your local library, preview in-app, and add audio to your restored videos — all without leaving ATR.
A built-in AI chat assistant turns plain English into a restoration pipeline — and everything the app does is scriptable from the command line.
Describe the job — “clean up this tape and make the faces sharper” — and the assistant plans the engine stack, renders preview candidates of real frames, and lets you compare them side by side and pick one before committing to a full render. No pipeline knowledge required.
AI AssistantBring your own API key: Anthropic Claude, OpenAI, or any OpenAI-compatible endpoint — including a local LLM on your own hardware for a completely offline, private setup. The assistant is optional; every feature works without it.
Your ChoiceThe packaged app ships atr-cli with every subcommand — probe sources, apply presets, create and run render jobs, batch overnight queues. Wire ATR into your own scripts, watch folders, or digitization workflow.
Assistant-guided tools that find the story in hours of raw footage — scene- and people-aware selection that builds highlight timelines and exports them as EDL, FCPXML, or DaVinci Resolve projects for finishing in your NLE.
Story ToolsOne click opens your restoration in DaVinci Resolve with matching color-managed project settings — wide-gamut mezzanine in, correct data levels, no guesswork. Finish grading where professionals finish.
NLEThree pipeline hook stages let plugins process frames before filtering, after filtering, and after encode — write your own per-frame Python filters or install community plugins from the built-in marketplace browser.
ExtensibleStart free with the open-source Community Edition, or buy Pro once and own it forever — your license covers all v6.x updates, including every new AI engine added during the v6.x lifecycle.
No AI engines or batch processing — those are Pro (v6.x).
Download Free on GitHubPrice increases automatically to $150 after sale #50
Contact [email protected]. Include your order ID if you have one — it gets you the fastest response.
Your license key is emailed automatically by Lemon Squeezy the moment your purchase completes. Check spam first — if it has not arrived, contact us with your order ID.
Always verify with the SHA256 checksum to confirm a safe, untampered installer.
Full product manual, quick start guide, capture guide, plugin development guide, and AI feature guides included with the app.