All Alchemy pages

Describe a picture

Reads an image and answers with words, tags or numbers instead of another image.

Each form takes the same source image and returns different kinds of responses: a one-line caption, a tag list, a safety verdict, a colour palette. Several can run on one submission, so a single upload can come back captioned, tagged and rated at once. Each operation is charged separately, but the horde runs them at the same time and returns them together.
Caption
A short, one-line description of the image.
Interrogation (tags)
A ranked list of descriptive tags, each with a confidence score.
NSFW check
A single yes/no flag for adult content, nothing else.
Palette
Extracts the image's dominant colours as a swatch palette.
Image Properties (describe)
Returns technical metadata: dimensions, format, aspect ratio, and perceptual hashes (dhash/phash) and other image properties.
Aesthetic score
Rates the image on a 0-10 aesthetic scale using the LAION aesthetic predictor. Higher is more visually appealing.

Image settings reference

Examples

Caption
The image that was submitted 1024 × 1024 Source
an older woman dressed in blue posing for a picture while overlooking mountains at sunset with her arm up
 Result

A short sentence describing the picture, from the BLIP captioner. The horde's alchemist workers serve this form, not anything local: the answer was asked for through the public API and is published as it came back.

Interrogation (tags)
The image that was submitted 1024 × 1024 Source
Tags 5
  • alicia-viewstream27.1%
  • aunt15.7%
  • grandmother4.6%
  • mon2.7%
  • mother1.9%
Mediums 5
  • renaissance painting12.7%
  • photorealistic painting10.0%
  • stock photo9.9%
  • poster8.9%
  • matte painting7.4%
Techniques 5
  • Anamorphic54.1%
  • Photorealism12.7%
  • Architectural4.2%
  • Geometric3.5%
  • Single Color2.6%
Movements 5
  • neo-romanticism12.3%
  • aestheticism7.6%
  • retrofuturism7.4%
  • german romanticism6.0%
  • panfuturism5.2%
Flavors 5
  • Bildungsroman16.0%
  • Atompunk12.5%
  • Mythpunk9.9%
  • Steelpunk4.0%
  • Solarpunk3.5%
Artists 5
  • Yael Bartana14.2%
  • Maria Graham6.8%
  • Annie Leibovitz5.1%
  • Margarete Schütte-Lihotzky4.1%
  • Mela Muter3.4%
Likely sources 5
  • behance59.0%
  • cgsociety21.7%
  • pinterest5.0%
  • Artstation3.5%
  • cg society2.1%
 Result

The tags and phrases CLIP thinks fit the picture, which is a prompt read back out of an image rather than a sentence about it. The horde's alchemist workers serve this form, not anything local: the answer was asked for through the public API and is published as it came back.

NSFW check
The image that was submitted 1024 × 1024 Source
No NSFW detected
 Result

Filed under describe because its answer is a judgement about the picture rather than a change to it; it is a safety check and not a description, so the fit is approximate. The horde's alchemist workers serve this form, not anything local: the answer was asked for through the public API and is published as it came back.

Image Properties (describe)
The image that was submitted 1024 × 1024 Source
Format
WEBP
Width
1024px
Height
1024px
Aspect ratio
1.0000
Alpha channel
No
dHash
d8d8c6c75b096aea
pHash
cfd6531a01ace4e5
BlurHash
LLD9el-oVrRj~W%MV?n#IpW=Rjxa
 Result

Dimensions, dominant colour, blurhash and perceptual hashes: what the file is rather than what the picture shows. The horde's alchemist workers serve this form, not anything local: the answer was asked for through the public API and is published as it came back.

Aesthetic score
The image that was submitted 1024 × 1024 Source
6.396/ 10

Good

Result

Filed under describe because it answers a question about the picture in a number; it describes taste rather than content, so the fit is approximate. The horde's alchemist workers serve this form, not anything local: the answer was asked for through the public API and is published as it came back.

Palette
The image that was submitted 1024 × 1024 Source
#0d1c1e23.9%
#39302821.9%
#8c8a8c17.4%
#35555915.6%
#786e6a11.2%
#d2c0af10.1%
 Result

Filed under describe because its answer is a read of the picture returned as data; it is a colour list rather than a description, so the fit is approximate. The horde's alchemist workers serve this form, not anything local: the answer was asked for through the public API and is published as it came back.

Open the Alchemy studio