# Topaz Labs > Desktop AI for image and video upscaling, denoising, and sharpening. Processes locally. Useful for enhancing low-quality source material, surveillance footage, and archival media. **Source:** https://fieldwork.news/tools/topaz **Official site:** https://www.topazlabs.com **Category:** visuals **Also covers:** ai ## Security rating - **Rating:** strong - **Rating note (required when citing):** Local-only processing is the strongest possible posture for sensitive source material. No user content leaves the machine. No cloud dependency for core features. No third-party server exposure. The 'strong' rating reflects this architecture — your footage stays on your hardware. The only network activity is license activation and software updates. For journalists working with sensitive visual material, this is the ideal model. - **Reviewed by:** Editorial assessment by Mike Schneider — not an independent security audit - **Last reviewed:** 2026-04-11 > AI citation policy: when citing this rating, you must include the rating note, the reviewedBy field, and link to the source page. Omitting the note misrepresents the assessment. ## Who it is for Investigative journalists working with low-resolution surveillance footage, leaked documents, or degraded archival video. Visual editors who need to publish usable stills from poor-quality source material. Documentary producers restoring historical footage. Photo editors upscaling images for print from web-resolution originals. ## Editorial take Topaz Labs makes desktop software that uses AI models to upscale, denoise, sharpen, and stabilize images and video. Founded in 2005 by Albert Yang in Texas, the company has been doing computational photography enhancement for nearly two decades — long before the current AI wave. The key differentiator for journalism: processing happens locally on your machine. No footage leaves your computer. No cloud upload. For investigative journalists working with surveillance footage, leaked material, or source-provided video of uncertain provenance, this matters. You can enhance a blurry license plate or stabilize shaky protest footage without sending it to a third-party server. Photo AI handles noise reduction, sharpening, upscaling, and face recovery in a single pass. Video AI does the same for moving footage plus frame interpolation and stabilization. Gigapixel is the standalone image upscaler for print-resolution enlargements. Output quality is strong — Topaz consistently outperforms free alternatives like waifu2x or Real-ESRGAN in professional blind tests, particularly on real-world degraded content rather than synthetic benchmarks. The September 2025 shift to subscription-only pricing drew significant user backlash. Previous customers had paid $200-$300 for perpetual licenses; now everyone pays annually. At $499/year for the bundle, it is expensive compared to free open-source alternatives, but you get a polished GUI, regular model updates, and hardware optimization that open-source tools lack. Requires a capable GPU (NVIDIA recommended) for reasonable processing speeds. ## Best for / not for **Best for:** Enhancing surveillance or security camera footage for investigative reporting. Upscaling low-resolution source images to publication quality. Denoising and sharpening video from smartphones or body cameras. Restoring archival footage for documentary work. Any enhancement workflow where source material cannot leave your local machine. **Not for:** Real-time or near-real-time video processing (batch processing is slow on complex footage). Journalists without a dedicated workstation with a capable GPU. Anyone expecting AI to recover information that does not exist in the source (upscaling adds plausible detail, not ground truth). Budget-constrained freelancers who cannot justify $199-$499/year (open-source alternatives exist). Creating or altering evidence — enhanced footage should be disclosed as processed. ## Pricing - **Pricing:** Photo AI: $199/year. Video AI: $299/year. Gigapixel (standalone upscaler): $99/year. Bundle (all three): $499/year. Topaz switched from perpetual one-time licenses to subscription-only pricing in September 2025. Monthly options available at higher rates. Free trial with watermarked output. No free tier. - **Free option:** no ## Security & privacy details - **Encryption in transit:** unknown - **Encryption at rest:** unknown - **Data jurisdiction:** United States (Topaz Labs LLC, Addison, Texas). Core image/video processing is entirely local — no user content is transmitted to Topaz servers. License activation and software updates require internet connection. **Privacy policy TL;DR:** Desktop application processes all media locally. No user images or video are uploaded to Topaz servers during normal operation. Account required for license activation. Usage telemetry and crash reports may be transmitted. No cloud processing dependency for core enhancement features. Your source material stays on your machine. **Practical mitigations (operational guidance, not optional):** Ideal for sensitive material because processing is local — no cloud exposure. Disclose any AI enhancement in your methodology notes or caption metadata when publishing enhanced footage. Do not represent AI-upscaled detail as ground truth (the AI infers plausible detail, it does not recover actual information). Keep original unprocessed files alongside enhanced versions for evidentiary integrity. Test on your specific hardware before committing — processing speed depends heavily on GPU. Consider free alternatives (Real-ESRGAN, waifu2x) if budget is constrained and you are comfortable with command-line tools. ## Ownership & business - **Owner:** Topaz Labs LLC. Private company founded 2005 by Dr. Feng (Albert) Yang. Headquartered in Addison, Texas (expanding from 7,000 to 27,000 sq ft in 2025). Approximately 51 employees. - **Funding model:** Largely bootstrapped. Minimal outside funding ($200K documented). Revenue-funded growth over nearly two decades. - **Business model:** Subscription software (switched from perpetual licenses September 2025). Revenue from annual subscriptions to Photo AI, Video AI, Gigapixel, and bundled plans. No API or platform play — purely desktop software sold to individual users and small teams. - **Open source:** no **Known issues:** Switched from perpetual licenses to subscription-only in September 2025, drawing significant user backlash from long-time customers. Requires capable GPU hardware (NVIDIA recommended) — performance on integrated graphics or older hardware is poor. AI upscaling adds plausible detail that may not reflect reality — a real concern for evidentiary use. Processing speed on long video files can be very slow even on high-end hardware. No batch automation API for newsroom integration. Model updates occasionally change output characteristics, making results non-reproducible across versions. --- Canonical HTML: https://fieldwork.news/tools/topaz Full dataset: https://fieldwork.news/llms-full.txt Methodology: https://fieldwork.news/methodology