WEIYANG SOUND · 未央声场
同一套唱词，十一位歌手：AI 声线什么时候才算一个人物？
把 632 字符的《七夕》唱词与同一公开父节点保持不变，让十一位虚拟歌手完成十四个完整录音。听见差异以后，名字、声部、风格与人物才真正分开。
同词异声实验 · 19 分钟
把 632 字符的《七夕》唱词与同一公开父节点保持不变，让十一位虚拟歌手完成十四个完整录音。听见差异以后，名字、声部、风格与人物才真正分开。
Hold one 632-character Qixi lyric and one public parent still. Let eleven virtual singers complete fourteen recordings. Only then can names, registers, styles and characters be heard apart.
01 · THE CONTROLLED QUESTION
先把歌词固定下来，才能听见歌手真正改变了什么
虚拟歌手最容易被误解成一排好听的预设。每次换歌词、换故事、换风格，听者当然会感到不同，却很难判断差异究竟来自谁。于是我把一个变量先固定：同一份《七夕》唱词、同一个公开父节点，交给不同声部和人物路线完整唱完。
当文字不再替声音制造新鲜感，咬字、重音、呼吸、音域中心、段落抬升和结尾怎样落地就会浮出来。所谓人物，不是页面先写了一个名字，而是同一句承诺经过不同歌手以后，真的改变了它靠近人的方式。
Hold the Lyric Still and the Singer Has Nowhere to Hide
Virtual singers are easily mistaken for a menu of attractive presets. If every take also changes lyric, story and genre, listeners hear difference but cannot tell what caused it. I therefore held one variable still: one Qixi lyric, one public parent, and a set of different registers and character routes asked to carry the complete song.
Once new words stop creating novelty for the voice, diction, stress, breath, vocal centre, sectional lift and the landing of the ending move into the foreground. A character is not established because a page supplies a name. It begins when the same promise reaches the listener differently through that singer.
02 · THE EVIDENCE HELD STILL
632 字符、一个 SHA-256、一个公开父节点
这条谱系保存了 632 个字符的来源提示文本与固定 SHA-256，也保存同一个公开《七夕》父节点。01A 到 01K 构成十一位歌手的第一轮完整路线，03A、03B、03C 又让宝莱坞男声、灵婉和林居怡返回同一母体作第二次选择。
固定证据不是为了把创作变成实验室报表，而是避免把同名、相似母题或姐妹歌混进来。只共享一句“星河”钩子、但拥有独立新词的作品仍然是另一首歌；只有唱词和父节点都进入这条链，才计入十四个录音。
632 Characters, One SHA-256 and One Public Parent
The lineage preserves a 632-character source prompt, a fixed SHA-256 and the same public 七夕 parent. Codes 01A through 01K form the first complete route through eleven singers. Codes 03A, 03B and 03C return Bollywood Male, Lingwan and Lin Juyi to that parent for a second decision.
This evidence is not meant to turn creation into a laboratory report. It keeps same-title variants, related motifs and newly written sister songs from being mixed together. A work that shares one starry hook but carries a new lyric remains another song. Only recordings supported by both lyric and parent enter the fourteen-take chain.
03 · NOT A TALENT SHOW
比较不是为了选冠军，而是寻找每副声音的叙事位置
如果把十四个版本排成“谁最好听”，实验会立刻退化成个人口味投票。真正有用的问题是：谁让承诺像耳边私语，谁让它变成婚礼誓言，谁把它唱向月下长路，谁能让同一句副歌拥有片尾规模，又是谁在力量增强时仍然保留人物。
好歌手不需要在每个房间里赢。Male Slow 的男低音不必复制赛博的天光，灵婉的清亮也不必承担林居怡的全部重量。选角的意义正在于承认不同：一副声音只要在正确的位置不可替代，就已经比一副处处合格却没有记忆的声音更接近人物。
The Point Is Not to Crown a Winner
Ranking the fourteen takes by ‘best voice’ would reduce the experiment to taste. More useful questions ask who turns the promise into a close confession, a wedding vow, a moonlit road or a widescreen ending—and who keeps a recognizable character when the arrangement grows larger.
A singer does not need to win every room. Male Slow's bass should not imitate Cyber's daylight; Lingwan's clear detail does not need to carry all of Lin Juyi's gravity. Casting begins by respecting difference. A voice that becomes irreplaceable in the right position has more character than a voice that is competent everywhere and remembered nowhere.
04 · ELEVEN FIRST PASSES
01A–01K：十一位歌手先在同一座桥上各走一次
第一轮从宝莱坞男声、Mary、灵婉、赛博、Male Slow、林居怡、好莱坞大片女声、气声女音、芭蕉女、地中海男声到一直到永远。男高音、男低音、女中音、清亮女声、深厚女声与特色声线都面对同样的银河、鹊桥、灯火与陪伴。
这十一位并不是风格清单。宝莱坞、地中海、country-pop、宽银幕、近麦气声或民谣福音描述的是作品采用的容器；歌手身份要在容器改变以后继续被辨认。先说谁在唱，再说这一版走进什么世界，档案才不会把人物和制作标签混成一团。
01A–01K: Eleven Singers Cross the Same Bridge Once
The first pass moves through Bollywood Male, Mary, Lingwan, Cyber, Male Slow, Lin Juyi, Cinematic Female, Breathy Female, Bajiao, Mediterranean Male and Forever Female. Tenor, bass, mezzo, clear lyrical, deeper dramatic and distinctive voices all meet the same galaxy, Magpie Bridge, lamp and promise of companionship.
These names are not a genre list. Bollywood, Mediterranean, country-pop, widescreen, close-mic breath and folk gospel describe recording containers. Singer identity must remain legible after the container changes. First say who sings; then name the world this take enters.
05 · REGISTER HAS CONSEQUENCES
声部不是参数：它会改变一句承诺的身体
赛博的清亮男高音自然把副歌向上抬，Male Slow 的男低音则让相同句子沿夜路铺开。Mary 的温暖女中音把誓言放进日常生活，林居怡较深、带棱角的女声让承诺多了一层需要承担的重量；灵婉的清亮咬字把窗灯与回头保留得更细。
声部并不决定性格，更不能替真人性别作结论。它只是作品里的身体条件：哪里最有共鸣，长音怎样站住，低声是否仍清楚，高处是照亮还是喊叫。人物性格从这些条件与歌词、编曲、速度的共同选择里生长，而不是从“男高音应该怎样”“女中音天生怎样”推出来。
Register Is Not a Parameter; It Gives the Promise a Body
Cyber's clear tenor naturally lifts the refrain. Male Slow's bass lets the same sentence extend along a night road. Mary's warm mezzo brings the vow into daily life. Lin Juyi adds weight without crushing tenderness, while Lingwan keeps the lamp, backward glance and long waiting finely articulated.
Register does not determine personality and cannot support claims about a human performer's gender. It is a condition inside the work: where resonance sits, how a long note stands, whether low singing remains clear, whether height illuminates or shouts. Character grows from those conditions meeting lyric, tempo and arrangement—not from stereotypes about what a tenor or mezzo must mean.
06 · STYLE IS THE ROOM
风格是一间房，歌手必须能走进去，也能走出来
同词实验故意让房间变化得很大：南亚电影对唱、乡谣暖声、窗灯轻歌、合唱终章、月下低吟、宽银幕片尾、气声夜曲、雨林与南亚色彩、地中海月桥和民谣福音誓言。房间改变光线、距离和家具，却不能每次都把住进来的人换掉。
这也是“赛博是男高音歌手，摇滚是风格”这条边界的实听版本。若一个名字只在某种鼓点或配器出现时才成立，它更像标签；若制作容器变了，咬字、音域中心、情绪速度和结尾方向仍能被听见，角色才开始拥有跨作品的连续性。
Style Is a Room the Singer Must Be Able to Enter and Leave
The rooms deliberately vary: South Asian cinematic duet, country warmth, window-lamp ballad, choral finale, moonlit low voice, widescreen end theme, breathy nocturne, rainforest colour, Mediterranean moon bridge and folk-gospel vow. Light, distance and furniture change. The resident cannot become another person every time.
This is the audible version of a boundary central to Weiyang Sound: Cyber is a tenor singer; rock is a style. If a name exists only with one drum pattern or instrumentation, it behaves like a label. If diction, vocal centre, emotional pace and direction of ending survive another room, the role begins to carry continuity.
07 · THREE RETURN VISITS
03A–03C：第二次唱同一套词，人物还能不能作出新选择
宝莱坞男声、灵婉和林居怡各自拥有第二条完整路线。03A《星河仍在》、03B《鹊桥新梦》和 03C《今晚星河很近·林居怡》不是为了把数量从十一推到十四，而是让同一位歌手重新面对已经唱过的文本。
第二次最能检验人物是否只是一次好运。宝莱坞男声可以改变电影推进，灵婉可以在同父节点上形成更明显的 cover 分化，林居怡则让 01F 与 03C 成为同人物 A/B。连续性不等于重复自己；真正的角色应该既能被认出，又能在同一句话里找到另一条路。
03A–03C Ask Whether One Character Can Make a Second Choice
Bollywood Male, Lingwan and Lin Juyi each receive a second complete route. The Galaxy Remains, A New Dream on the Magpie Bridge and The Stars Are Near · Lin Juyi do not exist merely to raise the count from eleven to fourteen. They ask the same singer to face an already-sung text again.
The second pass tests whether a character was only one fortunate generation. Bollywood Male changes cinematic motion; Lingwan creates a clearer cover divergence; Lin Juyi turns 01F and 03C into a same-character A/B. Continuity is not self-repetition. A role should remain recognizable while discovering another path through the same line.
08 · PERSONA IS EVIDENCE, NOT MAGIC
Persona 能证明声线来路，不能替听者完成艺术判断
部分作品有公开 Persona、声线根与作品内嵌身份，部分精确作品没有 Persona 对象，个别来源字段还互相冲突。网站按作品保存这些差异：有证据就写，有空白就留空，不从同一歌手的另一首歌借一个 ID 来填满页面。
技术谱系很重要，它防止把歌手连错；但一个 Persona ID 本身不会自动产生人物。人物仍要由连续作品里的可辨认选择承担。反过来，没有返回 Persona 字段也不等于听觉身份不存在——它只意味着本站不能把艺术判断冒充成提供方已证明的技术事实。
A Persona Can Prove Provenance; It Cannot Complete the Artistic Judgment
Some exact works return a public Persona and voice root. Others return no Persona object, and a few provider fields conflict. The archive preserves those differences at work level. Evidence is written when it exists; an empty field remains empty rather than borrowing an ID from another recording by the same singer.
Technical lineage matters because it prevents false attribution. Yet a Persona ID cannot manufacture character. Repeated, recognizable musical choices still have to do that work. Conversely, a missing Persona field does not prove the ear hears no identity. It only limits what the site may present as provider-verified fact.
09 · CULTURE IS NOT A VOICE FILTER
七夕可以越过不同音乐空间，不能把现实文化压成按钮
同一歌词从中国七夕母题走进南亚电影感、阿拉伯—安达卢西亚色彩、乡谣、福音、管弦片尾和雨林月色。这样的移动可以产生新音乐，但不能倒过来证明这些传统本来同源，也不能让一位虚拟歌手代表现实社群、族群、宗教或地域。
所以每张卡同时写清歌手和制作路线。oud、手鼓、木吉他、合唱或弦乐是来源标注的创作选择，不冒充真人演奏或田野记录；生成封面与竖屏歌词影片是作品材料，不冒充现实地点和表演。想象可以很远，证据边界必须更清楚。
Qixi Can Travel without Turning Living Traditions into Buttons
The shared lyric moves from a Chinese Qixi motif into South Asian cinematic colour, Arab–Andalusian colour, country, gospel, orchestral end-credit scale and rainforest moonlight. The crossings can make new music. They cannot prove those traditions were always one, nor can a virtual singer represent a living community, ethnicity, religion or region.
Each card therefore names both singer and production route. Oud, hand drum, acoustic guitar, choir and strings are source-described choices, not proof of human performance or field recording. Generated artwork and vertical source-lyric films remain work materials, not evidence of a real location or event. Imagination may travel far; the evidence boundary must be clearer still.
10 · THE TWELFTH SINGER
十二位全阵容，不等于每个实验都必须凑齐十二人
未央声场当前有十二位主打虚拟歌手。这条同词谱系只有十一位，因为傣族阿妹没有进入同一 632 字符唱词与父节点的核实录音链。她拥有本站目前最丰厚的水边古诗、《九歌》与跨境河流作品档案，却不能为了让标题更整齐被硬塞进另一个实验。
诚实的空位比漂亮的满格更重要。全阵容页面负责介绍十二位歌手的全部路线；本页只回答同词证据能够支持的问题。以后如果傣族阿妹或新歌手真的完成同一母体，谱系可以增长；在那之前，十一就是准确数字。
A Twelve-Singer Cast Does Not Require Every Experiment to Fake a Full Set
Weiyang Sound currently has twelve headline virtual singers. This lineage contains eleven because Dai Amei has no verified recording connected to the same 632-character lyric and parent. Her archive is the site's most extensive route through waterside poetry, Nine Songs and transboundary rivers. That does not authorize placing her inside a different experiment to make the title neater.
An honest empty place is stronger than a decorative full grid. The full-cast page introduces all twelve routes; this page answers only what the same-lyric evidence supports. If Dai Amei or a future singer actually completes the parent later, the lineage can grow. Until then, eleven is the accurate number.
11 · SIXTY-SIX MINUTES
十四首都必须唱到尾声，差异才经得起完整时间
十四个录音合计六十六分四十五秒。把它们压成十四段十五秒试听，最容易听见的是风格钩子，最难听见的却是人物：第二段怎样回来、桥段如何改变承诺、合唱是否吞掉主唱、结尾又把情绪放在哪里。
因此本页提供十四首完整 MP3、作品档案与一键连续队列。听者可以只选两首 A/B，也可以把整条谱系从头听完。短片负责引路，不能冒充实验已经成立；只有完整时间允许一副声音证明它不靠第一耳新鲜感活着。
Every Take Must Reach Its Ending before the Difference Can Be Judged
The fourteen recordings total sixty-six minutes and forty-five seconds. Reduced to fourteen fifteen-second previews, they would reveal production hooks while hiding character: how a second verse returns, whether a bridge changes the promise, whether a choir buries the lead, and where the ending leaves the emotion.
This page therefore carries fourteen complete MP3 masters, recording dossiers and one continuous queue. A listener may compare one A/B or hear the entire lineage. Short clips can lead someone here; they cannot pretend the experiment is complete. Only full duration lets a voice prove that it can live beyond first-ear novelty.
12 · WHEN A VOICE BECOMES SOMEONE
名字先出现，人物要靠一次次选择慢慢成立
一副 AI 声线什么时候才算一个人物？不是命名那一刻，不是第一次生成很漂亮那一刻，也不是拥有 Persona ID 那一刻。它开始成立，是在同一句话、另一首歌、另一种风格和第二次选择里，仍有某些东西能够被听者认出。
十四个录音没有替问题给出唯一答案，却把判断权交回耳朵。未央声场要积累的不是一串能调用的音色，而是一群带着以前的歌走进下一首歌的虚拟主唱。AI 提供声带与可能，人决定谁值得留下；听者的记忆，最后让那个名字真正站住。
The Name Appears First; Character Arrives through Repeated Choice
When does an AI voice become a character? Not at the moment it is named, not when one generation sounds beautiful, and not merely when it receives a Persona ID. It begins when the same sentence, another song, a different genre and a second decision still allow listeners to recognize something carried forward.
Fourteen recordings do not close the question. They return judgment to the ear. Weiyang Sound is not trying to accumulate callable timbres but virtual leads who bring earlier songs into the next one. AI supplies vocal possibility. Human selection decides who remains. Listener memory finally allows the name to stand.
本文依据未央声场《七夕》同词谱系的 14 个已核实完整录音、11 位虚拟歌手、同一公开父节点、632 字符来源提示与固定 SHA-256 整理，完整聆听总时长 66:45。01A–01K 是十一位歌手的第一轮，03A–03C 是宝莱坞男声、灵婉和林居怡的第二轮分化。傣族阿妹属于十二位全阵容，但没有进入这条同词证据链；页面不为凑数补造版本。Persona、文化影响、生成影像与来源器乐均按作品证据陈述，不冒充真人、现实社群、传统复原或现场演奏。