Cost Overview
Cost
Costs associated with creating and using Videlicet vary by archive. It is dependent on the quantity of data hosted, and the the accuracy of the transcription. Bonny Broom indirect costs such as overhead and G&A will also be factored into the end cost to the user.
Prompt Cost
You can find the cost per run after each prompt run. You will notice that Gemini Flash LLM is less expensive and will be the recommended LLM for prompts.
Estimated processing cost to translate a 20,000 page archive is:
Gemini Flash 2.5: $200
Gemini Pro 2.5: $600
Example #1: cursive transcription
- Using Gemini Flash 2.5, the cost to translate 1 page of cursive from the Maryland Court Records is $0.01, and the accuracy rate is 99%.
- Using Gemini Pro 2.5, the cost to translate 1 page of cursive from the Maryland Court Records is $0.03, and the accuracy rate is 99.5%.
The prompt used is: You are an expert in deciphering and transcribing historical legal documents, specifically court records from the 18th century United States. Your task is to transcribe the following document image to text, cleaning up errors or missed words in previous OCR attempts (included after the image description). Be sure to use only the given image and cropped images of sections within the text as hints for clarification. Find and correct any long s, ⟨ſ⟩. Use brackets to indicate any words you are not 100% sure of. At the top of the translation indicate the accuracy of the transcription. List corrections that were made during translation.
You can see the difference in cost and accuracy for each translation. I have specified in my prompt to indicate the accuracy of each translation for reference purposes.
Archive hosting cost
This is TBD..
Videlicet licence cost
This is TBD..
Quantifying AI costs with tokens
The token is the unit of data that an AI model will use to represent words and charactors. A token can consist of one letter, or short word may be represented by one token, and a longer word at least two tokens. Typically, there are between 500-1000 tokens per page of text, with 600 being the earmark. There are two types of tokens. Input tokens are the prompt that you pass to the AI engine. Output tokens are the text that is generated or the resulting text generated. Costs are different for each.
For example:
Per 1M tokens in USD:
Google Gemnini 2.5 Pro (best results, but expensive):
Input price: $1.25, prompts <= 200k tokens $2.50, prompts > 200k tokens
Output price: $10.00, prompts <= 200k tokens $15.00, prompts > 200k
Google Gemnini 2.5 Flash (fast, cheaper, slightly less quality):
Input price: $0.30 (text / image / video) $1.00 (audio)
Output price: $2.50
If you are processing 10 pages of text, the estimated costs will be: 10 pages x 600 tokens (est) = 6000 tokens (est)
Pro:$16.25 or $17.50
Input cost (short prompt):$1.25
Input cost (long prompt): $2.50
Output cost:$15.00
Flash: $2.80
Input cost: $0.30
Output cost: $2.50