From Dev.to Community
Tech Articles
Curated developer articles, tutorials, and guides — auto-updated hourly
orca_forgeSep 1, 2026 • 6 min read
Speech Speed Cannot Be Changed After Learning — Parameters Are Accepted but Ignored
The quirks in speech patterns and speech speed of a voice model are embedded in the corpus and canno...
#tts
0 0
orca_forgeSep 2, 2026 • 6 min read
A single rough clip can ruin the entire style — 5 ways to choose the right ones
If a rough clip is mixed in with the few clips used to register an emotional style, the synthesized ...
#tts
0 0
orca_forgeSep 3, 2026 • 11 min read
Code for preventing hallucinations only failed during hallucinations
The trimming process designed to remove hallucinated clips from the speech synthesis training corpus...
#tts#whisper
0 0
orca_forgeSep 3, 2026 • 10 min read
The '3-Character' Catchphrase That the Quality Gate Allowed Became the Model's Habit
A few hallucinations mixed into the training corpus became ingrained as a catchphrase in the voice m...
#tts
0 0
orca_forgeSep 3, 2026 • 10 min read
Dropping Candidates Due to Fixable Defects — A Story of Measuring Factory Flaws as Product Personality
In the selection of voice model candidates, we included 'resistance to generation breakdown' as an e...
#tts#mlops
0 0
orca_forgeAug 31, 2026 • 6 min read
Gacha for Voices — One Line of Caption and a Random Seed Bring Back the Same Voice Every Time
With TTS where a caption and a random seed determine the voice, recording that combination lets you ...
#tts#mlops
0 0
orca_forgeAug 31, 2026 • 6 min read
TTS Chosen for Audio Quality Was Too Slow for Conversations — Real-World Tests Show 2.5x Speed Difference Between 'Design' and 'Production'
We tried implementing a diffusion-based TTS system that generates voices from captions for conversat...
#tts#gpu
0 0
orca_forgeSep 2, 2026 • 6 min read
Where Did the AI Learn to Stretch Its Greetings Like 'Kon-nichiwaa'?
An AI voice model unexpectedly started elongating word endings, even though it wasn’t instructed to ...
#tts#whisper
0 0
orca_forgeSep 1, 2026 • 8 min read
The Tighter the Quality Gate, the More the Flat Reads Survive — Selection Bias from Verification
I auto-generated a training corpus with emotional expression and every take came out flat. The cause...
#tts
0 0
orca_forgeAug 31, 2026 • 3 min read
「ナレーターっぽい声」を24候補から機械に選ばせる — 耳で選ぶのをやめた日
役割に合う声を乱数種のガチャで引き当てるとき、全候補を聴いて選ぶのは続かない。話速・抑揚・安定性を自動計測して役割ごとに重み付けし、上位だけ聴く仕組みを作った話と、指標の限界で男性候補が全滅した話。
#tts
0 0
orca_forgeSep 2, 2026 • 6 min read
When '少々' becomes 'しょも' — Japanese characters removed by whitelist
A TTS preprocessing whitelist that removes emojis and symbols was also stripping out Japanese charac...
#tts#unicode
0 0
orca_forgeSep 1, 2026 • 6 min read
TTS that Changes 'Recording Location' Each Time It's Generated — Inconsistent Audio Quality Ruins Style
Even when using the same model and speaker, the frequency characteristics vary between clips. Traini...
#tts
0 0
Paxa LabsAug 28, 2026 • 3 min read
Introducing Paxa Labs
Paxa Labs is an AI research lab in Bangkok. We build speech, translation, and document models for...
#ai#tts#api#thai
5 2
MaheshSep 1, 2026 • 1 min read
How to Create Text to Speech with NeuTTS and Hermes Agent
Hermes agent pretty much rules the Text to Speech world with it's approach of supporting them out of...
#ai#tutorial#tts#agents
0 0
Great SageAug 29, 2026 • 2 min read
Qwen3-TTS vs Kokoro: which self-hosted OpenAI-compatible TTS to pick
A few weeks back I wrote up self-hosting Kokoro-82M as a CPU-only, no-GPU TTS option. Since then I'v...
#tts#opensource#selfhosted#api
0 0
Stanly ThomasSep 3, 2026 • 5 min read
Building an Audio-First Content Roadmap
Most content teams still treat audio as a bonus track. The article ships, someone remembers "we...
#audiofirst#contentstrategy#contentroadmap#tts
0 0
Norvik TechSep 3, 2026 • 4 min read
How 'Shōshō' Became 'Shomo'
Originally published at norvik.tech Introduction Explore the implications of the...
#webdev#tts#texttospeech#japanesecharacters
0 0