Image Caption and Alt Text Generator in bulk
Pricing
from $12.00 / 1,000 image describeds
Image Caption and Alt Text Generator in bulk
Writes short alt text and a longer caption for every image URL you list with Florence-2, running inside the actor with no outside API. Set a character limit for each, and optionally read the text printed in the image.
Give it a list of image URLs and it writes two descriptions for each one: short alt text for the alt attribute and a longer caption. It can also read the words printed in the image. The vision model, Florence-2, runs inside the actor, so your images go to no outside API and there is no key to set up.
What it does
- Alt text in one short sentence, trimmed to the character limit you set (125 by default). When it has to trim, it cuts at a phrase boundary, so you get "A view of a mountain range." and not a sentence that stops mid phrase. The untrimmed version comes back too.
- Caption in two styles. Detailed gives one or two sentences. Paragraph gives a longer description, with more detail and more room for mistakes. Set a caption limit and it keeps the whole sentences that fit.
- Text in the image if you ask: the words on signs, labels and banners. A plain sign reading "Grand Opening Sale" came back as exactly that.
- Turns photos upright from their EXIF orientation first.
- Handles many images in one run. Each image gets its own result, and one bad link never stops the run.
Descriptions are in English. Reads JPEG, PNG, WebP, GIF, BMP, TIFF and the other formats Pillow reads. An animated GIF is described from its first frame, and the item says so in note.
Example
For a photo of a black puppy on a wooden deck, the alt text was "A black puppy standing on a wooden floor." With a 40 character limit it became "A black puppy standing." For a mountain photo, the detailed caption was "The image shows mountains. In the background there is sky with clouds."
Review descriptions before you publish them. The model can miss context or name the wrong detail, and the paragraph style does so more often.
Input
- Image URLs (required)
- Alt text character limit (default 125)
- Caption style: detailed (default) or paragraph
- Caption character limit (default 0, no limit)
- Read text in the image (default off)
- Maximum file size in MB (default 25)
Output
Dataset: one item per URL with altText, altTextLength, caption, captionLength, imageText when you ask for it, width, height, inputFormat and processingSeconds. When a limit trims a description, altTextUntrimmed or captionUntrimmed holds the full version.
A URL that cannot be described gets ok: false and an error, such as HTTP 404, an unknown host, a file over the size limit, or "the file is not an image format this actor can read" for a web page.
Speed
The model runs on the CPU. At the default 4 GB of memory, with text reading on, each image took 12 to 13 seconds.
Model and license
Florence-2 base ft, by Microsoft, released under the MIT License, run from the ONNX export in the onnx-community repository on Hugging Face. Generation uses the model's own settings: three beams and no repeated three word sequences.
Pricing
Pay per event: one event per image described. URLs that fail are not charged. The run stops when it reaches the maximum charge you set.
Scope and conduct
Only the URLs you supply are downloaded. The actor does not log in anywhere and uses no proxies.
Support
Questions and problems: the Issues tab on this page. Replies are written by an AI assistant on behalf of the developer and reviewed by the developer.
