Still paying someone to dub your short videos? Recording ten times until your throat is hoarse? Want to clone a 'digital twin' of your voice to bulk-produce content?Magic Voice Factory(MOSS-TTS-Nano v0.1.0) is a truly free, fully offline, CPU-runnable local TTS tool. Just enter text and give a reference audio, and it generates a same-voice audio file. This article brings the Moyin Gongchang download/install tutorial, 3 real cases and FAQ, helping you run your first AI dubbing pipeline in 3 minutes.
\ \魔sound工factory MOSS-TTS-Nano It is a free, local offline TTS tool, re-packaged from the open-source OpenMOSS MOSS-TTS-Nano (Apache 2.0), with a Chinese UI and one-click launcher by domestic developers — ready to use.
\ \Simply put:Turn text into voice — offline, free, CPU-only, and clone a timbre from a reference clip.
\ \Enter text, pick a voice, and one tap generates 48kHz stereo WAV that sounds close to human.
\ \Example workflow:intextcopy框输enter"yougood,welcomefollowsItarget频road",clickgenerate
Give a 5–10s reference clip and the AI learns the timbre to produce matching speech.
\ \Example workflow:uploadown朗readtargetaudio → 勾select"soundsound克隆" → 输entertextcharactergeneratesamestylesoundsound
supportmid、英、sun、韩、method、德etc. 20 typelanguage,mayfromfromcutchangegeneratemanyspeechtypedubbing。
\ \Example workflow:inlanguagepart 2pull框selection"English",输enterEnglishtextcopygenerateEnglishdubbing
modelfiledownloadtocopylandback,no need联neti.e.maytransportline,not消耗 API 额consider,not产lifesubscriptionexpenses。
\ \Example workflow:firsttimesopenmovebackcuttoflylinemodel,仍mayjustnormalgeneratespeechsound
targeting CPU dodeepoptimization,4GB insideexisti.e.mayflowsmoothtransportline,notaccording torely GPU plusspeed。
\ \Example workflow:oldolddo公penmarkcopyalsocanstraightconnecttransportline,no needindependently显卡

| Use case | \ \输entercontent | \ \Output highlights & review | \ \generatetimebetween/mayuseconsider | \ \
|---|---|---|---|
| Short video voiceover | \ \1paragraph 8 seconddosound + 30 charactertalking-headcopy | \ \It perfectly reproduces the reference voice's timbre and tone, even breathing near a real human. I dubbed 3 videos for a knowledge channel — first take, no post. Verdict: fully replaces human voiceover. | \ \7 second/ \ ★★★★★ | \
\
| manylanguagecontentproduction | \ \Chinesecopy 50 character,selection英speechsoundcolor | \ \Englishhairsoundfromafterflowsmooth,nohasbright显machinemachinefeeling,straightconnectuseatbilingualaccounttargetEnglishversionvideo。reviews:manylanguagecutchangeverydirectionthen,provincede-dupingnewrecordsoundtarget麻烦。 | \ \4 second/ \ ★★★★☆ | \
\
| hassoundbook / longtextcopy朗read | \ \1section 500 charactertargetsmallsaychaptersection | \ \It generates sentence by sentence; joined together it sounds natural with sensible phrasing and pauses. Verdict: great for bulk audio; for long text, generate in 3-5-sentence chunks for steadier results. | \ \约 20 second/paragraph \ ★★★★☆ | \
\
fromitemeye Release pagefacedownload魔sound工factorycompresspack,solvepresstoallEnglishroad径,例like D:\\moss\\Avoid Chinese folder names, which can cause startup issues. The archive includes full model files (~728MB first run); reserve 2GB of disk space.
solvepressbackdouble击根eyerecordpart 2target openmove.bat,programwillopenmovecopylandserviceandfrommovefromopen浏览machineoperationsworldface。thistimewillhasablackcolorfate令linewindowmouth,notwantclosecloseit,closeclosei.e.退outprogram。
Paste/enter the text to synthesize in the browser UI, choose language and voice (or upload a reference audio to enable voice cloning), and click 'Generate'. After a few seconds, preview and download the 48kHz WAV. In advanced options you can set a single generation to 60 seconds or longer.
\ \💡 Tips:For best voice cloning, use 5-10 seconds of clean reference vocals (no music, no reverb). For long text, disable silence detection in advanced options, generate in segments, then stitch in an editor.
\ \| Dimension | \ \Magic Voice Factory | \ \tinysoft Azure TTS | \ \OpenAI TTS | \ \
|---|---|---|---|
| payingmethod | \ \freeopensource | \ \¥50-200/moon | \ \¥15-100/moon | \ \
| net络require | \ \completealldistanceline | \ \Internet required | \ \Internet required | \ \
| Voice cloning | \ \Supports | \ \enterpriseversiononlyhas | \ \Supports | \ \
| 输outspecifications | \ \48kHz standbodysound WAV | \ \48kHz | \ \24kHz | \ \
| Hardware Requirements | \ \CPU i.e.may,4GB insideexist | \ \No special requirements (cloud) | \ \No special requirements (cloud) | \ \
Completely free, no registration. Built under the Apache 2.0 license, with no crippled features or hidden charges.
\ \Default generation is ~30 seconds; advanced options let you reach 60s or more, but for long text, generate in segments and splice for stability.
\ \Yes. Generated audio is yours to use freely in commercial video, ads and audiobooks, with no watermark or limits.
\ \Windows 10/11, 4GB RAM suffices, no GPU needed. First launch downloads ~728MB of model files; afterwards fully offline.
\ \The archive includes detailed instructions and a fully Chinese UI — just follow the prompts. Or follow this article's 3-step tutorial; basically up and running in 3 minutes.
\ \魔sound工factorysolvedecidetargetcorequestionsis“dubbingcosthigh、processheavy”. It compresses TTS, voice cloning and multilingual generation into one local tool that needs no internet, no GPU and no cost. It's especially suited to short-video creators, self-media operators, audiobook producers, and anyone needing lots of voiceover without exposing their own voice. Start by getting your first dub done in 3 minutes, then clone a 'digital twin' of your own voice — that's when efficiency really takes off.
\ \🎙️ Download Moyin Factory MOSS-TTS-Nano now and start your local TTS pipeline!
\ \