MikuTools 最新文章：工具教程、产品更新、AI 工具实践和工程笔记。
SSML is the standard for telling a text-to-speech engine how to read your text. Half the modern providers ignore it. Here is what is actually implemented across the major TTS APIs in 2026 and what to use when SSML is not on the menu.
The text-to-speech tool ships dozens of preset voices across nine languages. Which one fits a tutorial? A meditation? An ad read? Here is a practical taxonomy of when to reach for which voice, and how to test fast.
MP3, Opus, AAC, FLAC, WAV, OGG, and PCM all come out of the same TTS endpoint. They are not interchangeable. Here is which format to ask for, and when reaching for the wrong one quietly costs you bandwidth, quality, or compatibility.
Whisper supports 99 languages on paper. In practice, the accuracy spread between Tier 1 and Tier 4 languages is large enough to change which workflows are feasible and which are not. Here is a field guide.
Auto-generated captions look like an accessibility solution and usually aren't. Here is the difference between live captions and accessible captions, why ADA and WCAG care, and what good practice looks like in 2026.
A practical recipe for taking a 60-minute recorded podcast episode through transcription to ready-to-publish show notes, chapters, and social cuts in roughly 20 minutes of working time.
Speaker labels look magical and fail in predictable ways. Here is a practical guide to when to enable them, how to set the min and max speaker counts, and how to read the results.
Every transcription provider quotes a sub-7% WER on their landing page. The number is real on the benchmark and irrelevant to your audio. Here is how to think about provider choice when the numbers are this slippery.
The prompt field in Whisper-style transcription APIs is the most under-documented setting in the panel. Here is exactly what it does, what it does not do, and how to write one that actually works.
A practical decision guide for choosing between plain text, JSON, SRT, VTT, and detailed JSON when transcribing audio. The wrong format is a project on rails to needing a rerun.
Hyper3D's Rodin Gen-2 is a 10B-parameter model that turns text or images into production-ready 3D assets. Quad mesh, PBR materials, HD textures.
An opinion piece on why browser-side image compression has quietly become the right default for client-bound work, where cloud compressors still earn their keep, and the honest limitations of staying local.