v1.3.4

Patch Notes:

• FIX: Added new required library for fixing json format errors returned from Gemini and prevent unintended progress failures
  • json-repair == 0.39.1

Full Changelog: v1.3.3...v1.3.4

v1.3.3

Patch Notes:

• FIX: Prevent Gemini from hallucinating and merging dialogs when it shouldn't:
  • Changed request temperature to 0
  • Improved system instruction
• FIX: Removed final delay after all dialogs have already been translated when using Pro versions
• Bump google-generativeai version to 0.8.4

Full Changelog: v1.3.2...v1.3.3

v1.3.2

Patch Notes:

• Changed default model to the newly released Gemini 2.0 Flash

Full Changelog: v1.3.1...v1.3.2

v1.3.1

Patch Notes:

• Small but important fix for handling non-English input (thanks @sjiampojamarn)

Full Changelog: v1.3.0...v1.3.1

v1.3.0

Patch Notes:

• NEW: Added support for second API key for additional quota (thanks @CevreMuhendisi)
  • Automatically switches to the second API when the first API reaches its quota limit
• Better error handling
• Decreased wait time betweeen requests for Pro models when free quota and a second API key are detected

Full Changelog: v1.2.6...v1.3.0

v1.2.6

Patch Notes:

• Removed leftover print message call
• Proper handle of Gemini blocked requests

Full Changelog: v1.2.5...v1.2.6

v1.2.5

Patch Notes:

• NEW: Added two new optional variables:

  • description: Additional description for the translation job
  • free_quota: Variable to indicate if you are using free quota limits or not (Defaults to True)

• Fix for hitting free quota too fast when using gemini-pro models

• Better handling of context between requests

Full Changelog: v1.2.1...v1.2.5

v1.2.1

Patch Notes:

• Fix a small breaking syntax error when applying right-to-left compatibility (Thanks to @CevreMuhendisi)

Full Changelog: v1.2.0...v1.2.1

v1.2.0

Patch Notes:

• NEW: Added support for right-to-left languages in the subtitle translation process. (Thanks to @Theamoz)

• Improved batch processing and error handling in subtitle translation logic.

  • Brought back the send_message function call that was replaced by generate_content on v1.1.4
  • Context will now be preserved between batches again

• Added fallback to prevent infinite loop in case Gemini returns unexpected responses.

Full Changelog: v1.1.4...v1.2.0

v1.1.4

Patch Notes:

• Changed Gemini function from start_chat to generate_content

  • Chat feature was being used for saving context from previous batch but that would occasionally cause issues with Gemini returning the wrong number of dialogs from input (Gemini is dumb sometimes)

  • generate_content is more reliable albeit at the cost of context saving between requests (dialogs that start in one batch and end in another might lose a bit of continuity).

• Cleaned up code with modular calls to remove repetitiveness

Full Changelog: v1.1.3...v1.1.4