I consider everything that comes after I color correct my shots "work," and I hate working. So I built a desktop app so I don't have to.

TL;DR: My app does free, private descriptions and can reverse lookup color-corrected images and video to their sources for metadata and telemetry. No Subscription.

Get 100 descriptions with all features enabled with a no-sign up free trial: https://viscriptix.com

Tutorials Videos (many more coming):

DJI Drone Tutorial (best for seeing workflow + benefits): https://youtu.be/3iv_cMkSEPM
Using Proxies: https://youtu.be/sayfoSWODds
General Info: https://youtu.be/EcfP_og5QsI

Here are its benefits:

-$39. One-time fee. No subscription. Updates included.
-100 Describe free Trial with all features enabled. Try it at no risk.
-Reverse lookup for edited deliverables. Given a finished clip—cropped, color-graded, and cut down from the original—my algorithm locates the source file amongst thousands, pinpoints the exact frame offset, and pulls its metadata and telemetry. This is the most sophisticated part of the software.
-Privacy & zero ongoing cost. Runs 100% offline/airplane mode with all features enabled, including detailed geolocation, landmarks, subtitle generation, and fully featured AI descriptions via Ollama.
-Advanced DJI drone footage processing. Uses telemetry and optical flow to determine camera movement, time of day (sunset, night, etc.), seasons, holidays, landmarks, and location. Completely offline with zero AI.
-AI prompt extraction. Reverse lookup will find your original prompts and incorporate them into descriptions—for both AI-generated images and video generated from AI images.
-Catalog your footage. Create a searchable offline database of your media, including locations, landmarks, subtitles, camera settings, and more.
-Built to save you time at every step. Once properly configured, it's meant to be click-once-and-walk-away while you go shoot stuff or generate more AI slop.
-Verification. If you want to verify things like lookup matches, I've included tools—like the compare browser—that make it very fast to do so.

Many of the features in this app were based on discussions I read on this forum.

First 50 people to use the discount code MSTOCKG get $10 off. Unlimited descriptions for $29.



TIPS:

For Free Descriptions: Just download Ollama, sign up for a free account, and use "gemma4:31b-cloud" or "qwen3.5:latest-cloud" for free, fast online descriptions. If you have a fast machine, use one of the local gemma4 or qwen3.5 models for offline descriptions instead. To download any of these, type "ollama pull modelname" in a terminal. These models go toe-to-toe with the latest pay-to-play models from Google and OpenAI. To use untested/new models, enable them by pressing F12 (developer menu) -> model list.

Use Gemini and OpenAI models if you want high quality and high speed at ridiculously affordable rates (as low as $0.004 per description).

If you don't know how to set up an API or Ollama server, just follow the instructions in "AI Preferences." It only takes a few minutes.

Use Proxies: If you're doing video lookups into large archives, proxies speed up everything. Before you describe anything, before you do any lookups, before you generate hash indexes—generate proxies (use the in-app proxy generator) for both your deliverable media AND your source media. Matching proxy deliverables to proxy source media is exponentially faster. My workflow is often matching 4K ProRes to 4K d_log source on 4TB drives, and the time saved by converting to proxies up front is huge.

Benchmarks (note: small sample size):






About me: https://www.imdb.com/name/nm6705341/

I'm a software engineer, a filmmaker, and a colorist. I spent ten years in the video game industry and another ten as a timelapse photographer and 2nd unit director on several worldwide documentaries. I have a stock footage portfolio with over 10,000 videos and images (mixed—real natural history and AI-generated) and have been selling stock footage since 2009. I'm currently working on a suite of products that will automate everything between the memory card and your audience.

Also, don't let my benchmarks deceive you. All models listed are perfectly competent.

My accuracy benchmark grades on a curve. If subject matter is correct, it starts nitpicking things like, how close the character/word counts are to the target. All of these models listed were able to generate subject accurate descriptions with character/word counts within acceptable ranges. Models that failed at subject accuracy and character/word counts were completely left out of the benchmarks.

The cost chart above is per 100 descriptions.

If you use Gemini, costs can be as low as $0.004 per description.

If you use Ollama, descriptions are free and local models like Gemma4:e4b can run on most machines.

Local Ollama model description times were left out of the timing chart because they are completely subjective to the users machine. With my desktop, I can do gemma4:e4b descriptions in seconds but gemma4:31b completely chokes. You can always use gemma4:31b-cloud on a free tier Ollama account and it is fairly quick.

Viscriptix is now $19. $19 to get all your image and video keywording and descriptions for free and 100% offline while incorporating all of the metadata from your source files.

Download free trial with all features and exports enabled here: https://viscriptix.com/

If you have drone shots, this incorporate telemetry to describe location, time-of-day and camera movement.
If you have phone shots, this will incorporate GPS, location, date, time-of-day and auto-generated subtitles with conversation summary.
If you have video or images made from Midjourney AI prompts, this will incorporate the original prompt.
It has an advanced algorithm for generating and prioritizing keyword order for Adobe stock.
It has a multi-step AI refinement process for generating descriptions that ensures accuracy while keeping token use low.
By default it avoids all multi-word keywords (but it is configurable).
Use any AI provider you want, you can pay for OpenAI/Gemini or generate your own descriptions through local models on Ollama.
Zero information gathering.
You can turn on Airplane mode and/or turn off your wifi and use your own local models. Includes full geolocation database that runs offline.

Version 1.1.5 is up that includes several new features and bug fixes:

-Per-archive filename-pattern filter for indexing
-Pond5 export: SpecifySource column + lens-angle keyword
-Stop infinite retry loop on Gemini daily-quota exhaustion
-Improved drone camera movement classification accuracy
-Clear settings on uninstall