Subtitle Corrector v3.0

Changelog:

• Almost all exceptions that can be returned from a request to GPT4 are now handled.
• A ratelimit exception will cause the program to resend the request after a small delay, delay is increased with each rate limit error.
• Timeouts are handled by resending the request until either it is successful or 10 timeout errors have occurred.
• Content violation exceptions are handled by asking the user to add context to their query so that the AI doesn't complain. (Sidenote: It's strange we've reached a point where we have to sometimes resolve errors through negotiation)

Full Changelog: v2.8...v3.0

Subtitle Corrector v2.8

Changelog:

1. Prompts are no longer hardcoded and are instead stored in .json format.
2. Prompts are automatically loaded from prompt directory and can be selected from the dropdown menu in the GUI.
3. New config.json defines a directory for storing prompt, which GPT model to use, and how many tokens to use per query.
4. config.json is automatically generated from default if not found.
5. Cost output is now dependent on model used for processing.

Full Changelog: v2.2...v2.8

Subtitle Corrector v2.2

This update fixes a few small bugs and improves the capability of the program in handling exceptions.

Subtitle Corrector v2.0

Subtitle Corrector v2.0

This version of Subtitle Corrector contains is almost entirely rewritten to be asynchronous.
Because of this, it can now process a subtitle multiple times faster than previously possible.

Important changes:

1. All queries are now sent asynchronously.
2. Estimated cost is displayed at the end of the program's runtime.
3. Program now displays when a prompt has been received and how much time elapsed between the request and response.
4. Program now no longer writes constantly to the raw_output file.
5. Modified Dutch to English prompt due to Dutch output.
6. Encoding now varies depending on user's operating system.
7. Exceptions are now raised if a query's finish_reason is not stop.
8. num_tokens() now using the cl100k_base token encoding used by gpt-4, rather than the one used by gpt-2.

Full Changelog: v1.9...v2.0

Subtitle Corrector v1.9

Fixed crash due to missing "self.".

Full Changelog: v1.8...v1.9

Subtitle Corrector v1.8

Subtitle Corrector v1.8

New features in this release:

1. Program will now automatically correct output which has an inconsistent sub block count by sending the query again.

Subtitle Corrector v1.5

Subtitle Corrector v1.5: Prompt switching

This release contains the following changes:

1. Ability to switch between different prompts like "Dutch to English", "English to Dutch", and also standard subtitle correction.
2. Output file is now named as such: corrected_$ORIGINAL_FILENAME

Possible bugs:

1. Desynchronisation when processing large subtitle files. Most likely occuring when GPT output is being re-encoded into .SRT.

Subtitle Corrector v1.0.0

First release of the GUI version of the subtitle corrector program for Windows.

Download the -TERM edition if you want the program to open the terminal for progress reporting and status information.

Subtitle Corrector Alpha 0.5 (Windows)

Alpha version of the Subtitle Corrector program for Windows.
Query splitting functionality not yet implemented.