This issue collects feedback for using MyVoiceTyping in AI Coding workflows.
Why this matters
A lot of AI coding work starts with long natural-language prompts:
- bug reproduction steps;
- refactor instructions;
- feature requirements;
- PR notes;
- code review comments;
- debugging thoughts while reading code.
These prompts are often easier to speak than type, especially in Chinese or Chinese-English mixed workflows. But raw ASR text is usually not enough: it needs punctuation, sentence splitting, technical-term preservation, and light cleanup before it is safe to send to an agent.
What we want to learn
Please share non-sensitive examples of:
- Spoken intent
- MyVoiceTyping actual output
- Expected output
- Which AI coding tool you pasted into, if any
- Claude Code
- Qwen Code
- opencode
- Cursor
- GitHub Copilot Chat
- other
- Whether the local polishing model:
- preserved your meaning;
- changed too much;
- lost technical terms;
- added wrong details;
- made the prompt easier to send.
Example format
Tool: Claude Code / Qwen Code / Cursor / other
Language: Chinese / Chinese-English mixed / English
Scenario: bug reproduction / refactor / PR note / requirements / other
Spoken intent:
"帮我把这个 websocket handler 改成 async iterator,同时保留 backpressure 逻辑。"
Actual output:
...
Expected output:
...
What went wrong or right:
...
Privacy reminder
Please do not paste private repo code, customer data, internal API names, real logs, secrets, or sensitive work text. Synthetic examples are enough.
Related docs
This issue collects feedback for using MyVoiceTyping in AI Coding workflows.
Why this matters
A lot of AI coding work starts with long natural-language prompts:
These prompts are often easier to speak than type, especially in Chinese or Chinese-English mixed workflows. But raw ASR text is usually not enough: it needs punctuation, sentence splitting, technical-term preservation, and light cleanup before it is safe to send to an agent.
What we want to learn
Please share non-sensitive examples of:
Example format
Privacy reminder
Please do not paste private repo code, customer data, internal API names, real logs, secrets, or sensitive work text. Synthetic examples are enough.
Related docs