Voiceover and Audio Layer for the Vietnamese Language Course
Voiceover and Audio Layer for the Vietnamese Language Course
For Vietnamese, audio is not a decorative feature, but a fundamental part of the learning material. Tones, vowel quality, and finals must be available in real sound.
Why You Cannot Rely Solely on browser speechSynthesis
Initial research showed that the availability of a Vietnamese voice in the browser cannot be guaranteed. Installing additional voices depends on the end vendor, can be inconvenient and unstable, and users cannot be forced to install system dictionaries.
In addition, a developer who is not a native speaker cannot guarantee pronunciation quality solely based on a successful synthesis.
Custom Audio Dictionary
The course working architecture must store pre-prepared voiceovers:
- learning unit (letter, syllable, word, phrase);
- audio generation or recording;
- file saving;
- model/voice/pronunciation variant metadata;
- verification status;
- multiple reuse.
The voiceover does not need to be regenerated on every playback.
Native Speaker Verification
For critical elements—especially tones and minimal phonetic contrasts—a native speaker verification stage is required. It is useful to provide states such as "generated", "requires review", "confirmed", "rejected", and the ability to replace a synthetic variant with a live recording.
Dialects
If necessary, the audio dictionary should allow storing multiple variants of a single unit: for example, northern and southern.
Current Technical Research
The main TTS pipeline is being researched separately; an OpenRouter limitation on certain minimal inputs like a single letter has also been discovered, so a local TTS server is being considered.
Links
The audio material should be directly linked to Phonetics, Tones, Dialects, and Teaching Methodology.
