SONANO / Blog
Voice technology
& creation
Practical reading for local AI audio. Choose a model, prepare a voice, and turn your material into dialogue, subtitles or an audiobook.
Featured article
Qwen3-TTS: choosing CustomVoice, VoiceDesign or Base
Preset speakers, a written voice description, or a reference recording: choose the right starting point for your next piece.
Read articleKeep reading
15 articles · Models / Creation / AgentsQwen3-TTS, VoxCPM2 and OmniVoice: three paths to speech generation
How is sound represented, and how is that representation generated? Separate discrete codes, continuous latents and diffusion to understand three model families.
Read articleWhat does a voice carry beyond the words?
The same sentence can change with emphasis, pauses and intonation. Listening to those layers changes how we understand and evaluate synthetic speech.
Read articleVoice identity and performance: what comes after cloning a timbre?
Sounding like a speaker and delivering a line appropriately are different judgments. Explore reference recordings, direction and character continuity.
Read articleFrom demo clips to long-form audio: a different test for speech models
A natural sentence is only part of a long production. Chapters, context, continuity and revisions change what we should listen for.
Read articleSound as a tool: audio workflows for AI Agents
From an interview recording to a two-voice program, explore discovery, reusable voices, revisions and the points where human judgment belongs.
Read articleVoice design, voice cloning and speech generation: how they fit together
Choose a voice, then give it new words. Understand the three concepts and how design models, the Voice Library and speech models work together in Sonano.
Read articleVoxCPM2: finding a voice for narration and characters
Understand voice design, reference-based cloning and language coverage, then test a short passage before starting a production.
Read articleQwen3-ASR or Whisper? Compare language, timing and editing
Choose a transcription route around your actual deliverable: a readable transcript, timed subtitles, or structured data for the next tool.
Read articleCreate SRT subtitles from audio or video on your Mac
Import media, choose a local model, review the text and timing, and export captions that match the original or edited video.
Read articleMake a local audiobook on Mac: chapters, review and export
Build a project from EPUB, TXT or Markdown, settle on a voice, and generate chapters with room to review and revise.
Read articleGive your AI Agent local voice tools, one verified job at a time
Connect through Sonano, discover ready models and voices, then finish a small task and check the delivered audio.
Read articleLocal or cloud TTS? Choose around the work you actually make
Compare where your material goes, device requirements, repeated generation and editing effort before choosing a speech workflow.
Read articlePrepare a better reference recording for voice cloning
Clean speech and a matching transcript make a useful starting point. Test a short passage before committing the voice to a whole project.
Read articleChoosing Whisper: include the time you spend editing
Compare model size, waiting time and corrections on a real recording. Then review the transcript before turning it into subtitles.
Read article