ACE Studio

Lyrics and a melody become singing, with pitch curves you can draw.

Model profile

Vendor

ACE Studio

Form

插件

Input

歌词 · MIDI · 文本

Output

人声轨 · 音频

Licence

声库与商用授权分开核对，虚拟形象与声音的使用范围以官方条款为准

Verified

2026-10-30

Related

VoiceVibratoTimbre

What it lets you finish

Hand it lyrics and a melody and get a sung vocal track. You edit timing and pitch syllable by syllable on a piano roll, draw vibrato and slides, and shape breath and intensity. The result can be a lead, a harmony or a stacked layer, exported as audio or used inside the session. It handles the voice only; the backing is your work or another tool's.

Inputs and outputs

Input is lyrics, a MIDI melody and note lengths, with some workflows taking text direction for delivery. Output is a vocal track, exportable dry or with effects. It exists as a standalone program and as a plugin, the plugin letting you revise while listening in the DAW. Language coverage and available voicebanks follow the official pages. The listening examples and instances below are references for this kind of output, not this model's generations.

Strengths

• Syllable-level alignment: timing, pitch and consonant placement are editable, which makes diction more controllable than with text-only models.
• Pitch curves are drawable, so vibrato depth and onset can differ per phrase and long notes need not be flat.
• Several voicebanks can be stacked into harmony, making lead-plus-backing design direct.
• Working as a plugin means changing one word does not require exporting back and forth.

Weaknesses and failure modes

• Voicebanks are finite: the timbre you want may not exist, commercial use is licensed separately, and use of a virtual persona is a further question.
• Mandarin polyphones and neutral tones are misread, so every character needs its reading checked; skip it and one wrong word surfaces in the line. A reproducible test is a phrase full of polyphonic characters, heard as delivered by default.
• Long notes stay flat in expression unless you draw the swells, and otherwise sound like an organ rather than a person.
• Breath marks are inserted by rule, often between the wrong characters in fast songs, and need moving phrase by phrase.
• It produces voice only; arrangement, mixing and mastering remain yours, so treating it as a whole-song tool disappoints.
• After program or voicebank updates, vocals in old projects can change character, so export dry stems for anything that matters.

Terms, price and alternatives

Confirm: the commercial scope of each voicebank, limits on virtual personas and voices, what happens to exported audio after a subscription ends, and the official statement on training data and voice sources. Pricing is by subscription or by voicebank; exact figures and regional availability belong on the official page. Alternatives: DiffSinger for an open-source route with your own trained voicebank; Suno when you want a whole song with backing. This is the tool for work where every syllable of the vocal must be controllable.

Sources

• 产品形态、声库与编辑方式，依官方公开说明与条款页核对，未抄原文
• 失败模式依公开用户报告与创作实践归纳，未引用具体评测数据

Used for fact-checking only; all prose is original.