Use of OpenAI GPT models in translate5
The OpenAI plug-in makes it possible to use an OpenAI GPT model as a language resource for machine translation; either for pre-translation in projects or for usage in InstantTranslate.
The OpenAI GPT models can be created for specific language combinations and then fine-tuned using various parameters and, in some cases, individually trained.
Setting everything up
After activating the OpenAI plug-in, the following settings must be made in the translate5 system configuration:
OpenAI API URL (runtimeOptions.plugins.OpenAI.server)
This must read as follows: https://api.openai.com
Authentication token of the OpenAI URL (runtimeOptions.plugins.OpenAI.apiToken)
If your OpenAI account is authorized to use the API, you can create API keys via this page: https://platform.openai.com/api-keys
This page provides information on the best way to handle the API key: https://help.openai.com/en/articles/4936850-where-do-i-find-my-api-key
| Make sure that the payment options are filled in correctly. Missing payment information/settings can result in requests being rejected due to token limit overruns. In particular, the “auto recharge” option must be active so that the API can be used without interruption, see https://platform.openai.com/account/billing/overviewandhttps://platform.openai.com/account/limits. |
|---|
OpenAI API Organization ID (runtimeOptions.plugins.OpenAI.organisationId, optional)
If you use multiple OpenAI accounts and/or organizations, these IDs can be used to differentiate between them for billing and for token limits. This website explains how this can be set up and managed: https://platform.openai.com/account/organization
Standard match rate for machine translation with OpenAI (OpenAIMachine Translation default matchrate)
The automatic pre-translation is set to a match value of 70% or less by default. This value can be customized specifically for OpenAI if required.
