You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Advanced voice mode is a new kind of conversation mode in ChatGPT. Its structure seems to be confusing the export code, perhaps around multi-modal export.
Advanced voice mode conversations are generated on mobile, but then accessible on the desktop web version and in the official export. I am exporting all messages using the desktop web version (in Chrome), which is where Exporter hangs.
Bug Report
What happens: When I try to export all conversations (to either markdown or JSON), the export hangs when it encounters an advanced voice mode conversation. I think the same thing happens when trying to export a single advanced voice mode conversation, but I didn't diagnose that situation as thoroughly.
Expected outcome: Advanced voice mode conversations export normally, perhaps without the WAV file data from ChatGPT. (The contents of the WAV audio file are also provide as a text transcript.) This would allow all conversations to be exported.
As an interim outcome, it would be okay for Exporter to skip over advanced voice mode conversations, with a warning message, so the other messages are exported.
Workaround: Until this bug is fixed, I am identifying the advanced voice mode conversations and excluding them from the export. But it's time-consuming to identify and exclude them, and then they are not represented in the export.
Error Message
The following error message is displayed over and over in the console when processing an advanced voice mode conversation. Note, "Helpful Situation Explained" is the conversation name.
Request Helpful Situation Explained failed: TypeError: Cannot read properties of undefined
(reading 'startsWith')
at isMultiModalInputImage (userscript.html?name...fb3988e848a: 1178:61)
at Array filter (<anonymous>)
at userscript.html?name...fb3988848a: 1183:42
at Array. flatMap (<anonymous>)
at replaceImageAssets (userscript.html?name...fb3988848a: 1180:61)
at fetchConversation (userscript.html?name...fb3988848a: 1224:13)
Attachments
advanced-voice-demo-conversation-excerpted-2024-10-12.json - JSON excerpt showing a complete advanced voice conversation from the official export file (conversations.json). Note: Hex IDs (UUIDs and sediment IDs) were rewritten with anonymized IDs, with care taken to anonymize deterministically so references weren't broken. Apologies if there are any mistakes made in anonymization.
See pionxzh#261 for the situation. This commit avoids errors caused by multimodal parts we don't understand.
It does NOT add any functionality that understands advanced voice mode parts. Perhaps that should be broken out of pionxzh#261 into a separate enhancement request.
* work on #261: avoid hang
See #261 for the situation. This commit avoids errors caused by multimodal parts we don't understand.
It does NOT add any functionality that understands advanced voice mode parts. Perhaps that should be broken out of #261 into a separate enhancement request.
* address es-lint complaint
eslint error:
error '&&' should be placed at the beginning of the line
solution:
move a trailing `&&` to the front of the next line
Thank You
Thank you for ChatGPT Exporter. It is awesome!
Background
Advanced voice mode is a new kind of conversation mode in ChatGPT. Its structure seems to be confusing the export code, perhaps around multi-modal export.
Advanced voice mode conversations are generated on mobile, but then accessible on the desktop web version and in the official export. I am exporting all messages using the desktop web version (in Chrome), which is where Exporter hangs.
Bug Report
What happens: When I try to export all conversations (to either markdown or JSON), the export hangs when it encounters an advanced voice mode conversation. I think the same thing happens when trying to export a single advanced voice mode conversation, but I didn't diagnose that situation as thoroughly.
Expected outcome: Advanced voice mode conversations export normally, perhaps without the WAV file data from ChatGPT. (The contents of the WAV audio file are also provide as a text transcript.) This would allow all conversations to be exported.
As an interim outcome, it would be okay for Exporter to skip over advanced voice mode conversations, with a warning message, so the other messages are exported.
Workaround: Until this bug is fixed, I am identifying the advanced voice mode conversations and excluding them from the export. But it's time-consuming to identify and exclude them, and then they are not represented in the export.
Error Message
The following error message is displayed over and over in the console when processing an advanced voice mode conversation. Note, "Helpful Situation Explained" is the conversation name.
Attachments
Environment
ChatGPT Exporter:
Tampermonkey® 5.1.1
Chrome Version 129.0.6668.90 (Official Build) (arm64)
macOS Sonoma 14.7
MacBook Pro 2021, Apple M1 Max CPU
The text was updated successfully, but these errors were encountered: