gexgd0419
Random Letters Appearing in Microsoft Word When SAPI Voice Adapter Is Installed
Description: When the SAPI voice adapter is installed, random letters appear in Microsoft Word documents upon opening and reading them with a screen reader, even without making any edits.
Steps to Reproduce:
- Install the SAPI voice adapter.
- Open a local Microsoft Word document.
- Without making any edits, try reading the document with a screen reader (test with NVDA, which is free).
- Observe whether random letters appear.
- Close the document and reopen it—letters may appear in different places.
- Repeat the same test with a document stored on OneDrive.
Example of the Issue (Observed in a Real Case):
Original Document Text:
"The intent of both Pythagoras and his Egyptian and Babylonian predecessors was actually to gain an understanding of the surrounding world and find the most efficient way to interact with it, and their exceptional discoveries such as irrational numbers or fractions are therefore ascribable to the solution of everyday issues, such as the calculation of the salary due to the workers employed in the construction of pyramids or ziggurats."
After Opening and Reading with a Screen Reader:
"bbbbbbbbbbbbbbbbbbbbbbbbThe intent of both Pythagoras and his Egyptian and Babylonian predecessors was actually to gain an understanding of the surrounding world and find the most efficient way to interact with it, and their exceptional discoveries such as irrational numbers or fractions are therefore ascribable to the solution of everyday issues, such as the calculation of the salary due to the workers employed in the construction of pyramids or ziggurats."
The extra letters appear randomly and may change position each time the document is reopened.
Additional Details:
- The issue occurs even without editing the document—simply opening and reading it is enough.
- OneDrive does not seem to be the cause, as the issue happens with both local and cloud-stored files.
- Screen readers (JAWS/NVDA) are likely not the cause, as the issue persists regardless of the voice used (JAWS voices, Microsoft neural voices, etc.).
- The issue disappears when the SAPI voice adapter is uninstalled.
System Information:
- Operating System: Windows 11 24H2 (OS build 26100.3194)
- Microsoft Word Version: Microsoft Word Subscription Version 16.0.18429.20158
- Screen Reader: NVDA 2024.4.1 and JAWS professional 2025.2412.50
- SAPI Voice Adapter Version: 0.2.3
Expected Behavior: Word documents should display normally without extra letters appearing.
Actual Behavior: Random sequence of letters appear in documents upon opening and reading them with a screen reader. The letters may move to different positions each time the document is reopened.
Notes:
This issue needs further testing both with local and OneDrive files, to determine the root cause.
Does this still happen with the latest version of NVDA, and the latest version of NaturalVoiceSAPIAdapter?
If this still happens, please provide the log file of NVDA and NaturalVoiceSAPIAdapter, with the log level set to debug.
Thank you for your support and follow-up.
I have not yet installed NaturalSapiAdapter on my machine because I was concerned it might be causing the problem with random letters appearing in Word documents. However, after careful investigation, it appears that the root cause is more likely ABBYY FineReader 16, which I have uninstalled. Since then, the problem seems to have disappeared.
I will proceed with installing NaturalSapiAdapter on a freshly formatted PC and perform thorough testing to see if the issue reoccurs. I will provide feedback and logs as needed.
Separately, I would like to discuss a related feature request regarding language tags in Word documents and their support in screen readers using NaturalSapiAdapter
When a text document contains, for instance, mostly English but includes segments tagged as German using Word’s proofing language feature, the ideal behavior is for the screen reader to automatically switch voice according to these tags. In my testing, using screen readers like JAWS and NVDA with their native voices, the voice switches properly to German for those tagged segments, with brief pauses before and after. Using non-multilingual voices via NaturalSapiAdapter, the voice also switches language correctly, although with longer pauses. However, with multilingual voices via NaturalSapiAdapter, the reading tends to use the dominant language voice for the entire passage—ignoring the language tags—and does not switch dynamically, leading to mispronunciations.
This leads me to ask if it could be possible to enhance NaturalSapiAdapter to respect Word’s proofing language tags even when using multilingual voices, enabling seamless and accurate pronunciation changes within mixed-language texts.
I appreciate your attention and look forward to testing NaturalSapiAdapter and collaborating on this enhancement.
Best regards.
On Fri, 8 Aug 2025 at 13:05, gexgd0419 @.***> wrote:
gexgd0419 left a comment (gexgd0419/NaturalVoiceSAPIAdapter#42) https://github.com/gexgd0419/NaturalVoiceSAPIAdapter/issues/42#issuecomment-3167495303
Does this still happen with the latest version of NVDA https://www.nvaccess.org/download/, and the latest version of NaturalVoiceSAPIAdapter https://github.com/gexgd0419/NaturalVoiceSAPIAdapter/releases/latest?
If this still happens, please provide the log file of NVDA and NaturalVoiceSAPIAdapter, with the log level set to debug.
— Reply to this email directly, view it on GitHub https://github.com/gexgd0419/NaturalVoiceSAPIAdapter/issues/42#issuecomment-3167495303, or unsubscribe https://github.com/notifications/unsubscribe-auth/BPSGJGWFFQKNHEW3IZV3ADT3MR767AVCNFSM6AAAAABXH5EZZ2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTCNRXGQ4TKMZQGM . You are receiving this because you authored the thread.Message ID: @.***>