What Is Stills Lab? A Free Film-Still Database for AI Prompting
Stills Lab is a free film-still database with color palettes, lens data, and visual search, built to help AI filmmakers write better prompts.
What is Stills Lab?
Stills Lab is a searchable database of still frames pulled from films, TV shows, and music videos, built as a reference tool for anyone prompting AI image or video models. Each still comes attached with metadata: the color palette with hex codes, the lens used, and the camera used. That combination of visual reference and technical data makes it useful for people trying to get a specific cinematic look out of tools like Nano Banana, GPT Image, or any video generator that responds to detailed prompts.
TL;DR
- Stills Lab pairs film stills with technical metadata, including hex color codes, lens information, and camera details for each frame, which gives prompt writers concrete language to describe a look instead of guessing at adjectives.
- A visual search feature lets you upload an image and find stills with a similar tone or palette, a capability not commonly found in comparable film-reference tools.
- The tool sits in the same category as ShotDeck, an existing film-still reference site, but the metadata depth and visual search set it apart.
- A common workflow is screenshotting a still’s metadata panel and feeding it to an LLM (inside a canvas tool or elsewhere) to convert the color and lens data into a usable image prompt.
- The goal isn’t to recreate a shot exactly but to land “in the ballpark” of its look, tone, and color grade.
- Access is free, with a paid tier around $4 a month for expanded use, which is low cost relative to the value for anyone doing regular AI film work.
Seven tools to build an app. Or just Remy.
Editor, preview, AI agents, deploy — all in one tab. Nothing to install.
Why does color and lens metadata matter for AI prompts?
Image and video models respond much more reliably to specific technical language than to vague mood words. Telling a model to make something “moody” or “cinematic” is open to wide interpretation. Giving it an actual hex code for the shadows, a named lens (which implies a certain depth of field and distortion), and a camera body (which implies grain, sharpness, or dynamic range characteristics) gives the model concrete anchors to work from.
This is the core reason a resource like Stills Lab has value beyond being a mood board. Anyone can screenshot a nice-looking frame from a movie. Fewer people have access to the hex codes of its color grade or a plausible answer for what lens created its background blur. By surfacing that metadata directly on each still, Stills Lab turns a visual reference into something closer to a technical spec sheet, which is exactly the kind of input prompt-based generation models handle well.
How do you actually use Stills Lab to build a prompt?
The straightforward path is to browse or search for a still with the tone you’re after, then hand its metadata to an image model directly: drop the still into GPT Image or Nano Banana and manipulate from there using the color and lens information as guidance.
A more deliberate workflow involves an intermediate step through a large language model. The process looks like this:
- Find a still in Stills Lab with the right look and open its detail panel.
- Screenshot the panel, which includes the hex codes, lens, and camera data alongside the image.
- Drop that screenshot into a canvas-style tool or any interface connected to an LLM.
- Prompt the LLM to extract the color information and camera details and turn them into a usable image-generation prompt.
- Take that generated prompt into an image model (or video model) of choice to produce a new image with a similar visual signature.
The output won’t be a one-to-one match to the original still, and that’s not really the point. The goal is a shot that shares the same color temperature, lighting logic, and lens character as the reference, applied to entirely new subject matter. This distinction matters: using Stills Lab well means treating a reference frame as a technical starting point, not a template to copy.
Another use case is a color pass on an existing image. Take an image you’ve already generated, pair it with a Stills Lab reference still, and run both through an LLM or image editor to shift the new image’s grade toward the reference’s palette.
What makes the visual search feature different?
Most film-still libraries rely on keyword search: title, director, genre, or shot type. Stills Lab includes a visual search function that works the other way around. Upload an image, and it returns stills from its library that share a similar visual tone, even across completely unrelated films.
Remy doesn't write the code. It manages the agents who do.
Remy runs the project. The specialists do the work. You work with the PM, not the implementers.
In practice, that means feeding in an original AI-generated image, say a character in a diner, and getting back stills from films with a matching color mood, regardless of genre or era. The connections aren’t always obvious on the surface (a period drama and a modern remake showing up together because they share a warm, desaturated palette), but that’s the point. It surfaces reference material a person browsing by title or keyword would likely never find on their own.
This also doubles as a discovery tool. Scrolling through visual search results turns up stills from films and shows outside the usual recommendation algorithms, some obscure, some foreign, some music videos most viewers wouldn’t encounter otherwise.
Is Stills Lab worth using over other film-reference tools?
For anyone already using a tool like ShotDeck for cinematic reference, Stills Lab covers similar ground: a browsable archive of film and TV stills organized for creative reference. What distinguishes it is the depth of metadata attached to each frame and the visual search function layered on top.
If your workflow is purely browsing for inspiration, either type of tool works. If your workflow involves actually constructing prompts for image or video models, the hex codes, lens data, and camera details give you material you can hand directly to an LLM or paste straight into a prompt, which shortcuts a lot of the guesswork usually involved in describing a “look” in words.
The free tier is enough to explore the library and test the workflow. A low-cost paid tier exists for people who want expanded access, and the price point is low enough that testing it costs little beyond time.
Frequently Asked Questions
What is Stills Lab used for?
It’s used as a visual and technical reference library for people building prompts for AI image and video generation. Each still includes color palette data, lens information, and camera details that can be translated into prompt language.
Is Stills Lab free?
There’s a free tier that gives access to the core library and search features. A paid tier exists for roughly $4 a month for expanded access.
How is Stills Lab different from ShotDeck?
Both are film-still reference libraries, but Stills Lab attaches deeper technical metadata (hex color codes, lens, camera) to each still and includes a visual search feature that finds stills with a similar tone to an uploaded image, rather than relying only on keyword or title search.
Can Stills Lab recreate an exact shot from a movie?
Not really, and that’s not the intended use. It’s better suited to landing in the general visual neighborhood of a reference shot, matching color grade, lighting, and lens character, rather than reproducing the original frame exactly.
Do you need to know cinematography terms to use it?
No. The metadata panel supplies the technical terms (lens names, camera models, hex codes) directly, so you can copy that information into a prompt or feed it to an LLM without needing prior knowledge of cinematography.
