More room to create.
No generation quotas.

There are 42 voice choices: 9 fixed voices and 33 voice-design presets, each with a stored audition, plus 20 creation templates. Character voices use a chosen sample for subsequent lines and require the character and reference packages. Website auditions are generated examples; some scripts are newer than those bundled with v0.4.6. Actual results depend on your local test and chosen voice sample.

Custom narration, projects and CLI creation require a purchased, activated license. Once activated, there is no generation-count quota. Text limits and hardware constraints still apply. Before buying, listen to the website samples and run the app’s compatibility checks and fixed test samples. Fixed tests cannot use your own script or export customer work.

Yes. Projects import TXT, CSV, TSV, DOCX and PPTX scripts and organize them by chapter, role and paragraph. Generate or redo individual lines and compare previous takes. DOCX extracts body text and tables. PPTX reads speaker notes first, falling back to slide text; there is no OCR or PDF import. Quick tasks and CLI batches support up to 5,000 characters and 100 segments; CLI segments are limited to 600 characters each. Projects can store up to 10,000 lines and 500,000 characters, splitting long lines into short segments for sequential generation. Organize long courses by chapter.

The initial setup downloads runtime components and voice packages. After preparation, narration, reference-voice reuse and license checks run locally and can work offline. Shengyue does not upload your scripts, reference recordings or finished audio.

Installers are available for Apple Silicon and Intel Macs running macOS 14+. Windows requires x64 and Windows 10 2004 (19041) or later; Windows 11 is recommended. Light mode needs at least 8 GB RAM; standard, character and reference modes need at least 16 GB, subject to the app’s checks. Intel Mac and Windows builds use CPU inference without CUDA; full acceptance testing on those platforms is still pending. Disk needs vary by platform, packages and existing downloads. Check the app before buying. Linux and Windows ARM are not supported.

Describe a character’s voice in words, record a sample, or select 3–30 seconds of clear speech you have permission to use and provide its exact transcript. An optional local transcription package of about 486 MB can help fill in the transcript; review it manually. Character voices use a chosen sample for future lines and retakes. Redefining a voice or using free-form design may change its sound. Standard preset voices support expression per line; light mode supports natural delivery only. In the current download, locked and reference voices retain their sample’s expression and do not add line-level emotion control.

Compare takes, trim waveforms, adjust volume and add fades while preserving the original audio. Projects can import background music, loop it and lower it during speech. Preview the first 30 seconds of the mix, then export merged WAV, MP3, SRT and ZIP. A finished export supports up to 5,000 audio segments and six hours; export longer projects by chapter. SRT timing follows generated segments, not word-level recognition or video alignment. The CLI’s regular ZIP export packages individual audio files without project mixing or subtitles.

Agents that can execute local commands, such as Codex, Claude Code and WorkBuddy, can call the CLI. The local model generates audio and returns file paths rather than embedding the audio in the conversation. Agent writing and reasoning remain subject to the provider’s charges; token use depends on the workflow. External agents handle scripts according to their own settings. There is no dedicated MCP server. The installer includes the CLI, and creation requires activation.

Download the app and run its compatibility checks and fixed test samples. Once quality and speed are suitable, purchase through the seller that supplied the software. Send the device code shown in the app and your order details to receive a device-bound SY2 activation code, which can be verified offline. Price, device count, updates and device-transfer terms depend on your purchase agreement. Contact the original seller for a replacement code. For purchase, activation or device-change questions, email freeteamlite@gmail.com. This website does not currently offer online checkout.