词曲分离
- 把完整混音拆成独立音轨，异步任务
- 成品是一个 zip 打包文件，内含人声与伴奏等分轨
- 用
clip_id引用源音频
认证
All endpoints authenticate with a Bearer token.
Get an API key: open the console and create one.
Add it to the request header:
Authorization: Bearer YOUR_OAIRoute_KEY请求参数
固定填 flowmusic，大小写不敏感。
待分离音轨的 clip_id，取自某个已成功任务的 result.music[].clip_id。
本接口没有其他可选参数，也没有长度或格式的额外校验规则。
Always flowmusic, case-insensitive.
The clip_id of the audio to separate, taken from result.music[].clip_id of a successful task.
There are no other parameters, and no extra length or format validation.
提交响应
状态码，成功为 200。
初始提交时为 submitted。
任务唯一标识，用于查询进度与结果。
The status code; 200 on success.
submitted at initial submission.
The unique task identifier, used to query progress and results.
结果字段
新结果的 clip_id。
分轨包下载地址。zip 格式。
与 file_url 相同的地址，两个字段指向同一份文件。
固定为 application/zip。
文件大小，单位字节。分轨包通常有 20MB 左右。
The clip_id of the new result.
The stem package download URL, in zip format.
The same address as file_url; the two fields point at the same file.
Always application/zip.
File size in bytes. Stem packages are usually around 20MB.
怎么用这个结果
返回的不是单条音频，而是一个 zip。下载后解压，里面按轨道分开：
| 文件 | 内容 |
|---|---|
| 人声轨 | 只含演唱部分 |
| 伴奏轨 | 去掉人声的器乐 |
| 其他轨 | 视模型能力，可能有鼓、贝斯等 |
拿到分轨后可以单独处理某条轨道，比如给人声加混响、替换伴奏，再重新合成。
解压前先确认 size_bytes 与实际下载大小一致。分轨包较大，网络中断时容易拿到残缺的 zip，解压会报错。建议下载后校验文件完整性。
What comes back isn't a single audio file but a zip. Download and unpack it, and the tracks are separated inside:
| File | Contents |
|---|---|
| Vocal track | Only the sung parts |
| Instrumental track | The music with vocals removed |
| Other tracks | Depending on model capability, possibly drums, bass, etc. |
Once you have the stems you can process one track on its own — add reverb to the vocals, swap the instrumental — and re-mix.
Before unpacking, confirm size_bytes matches the actual downloaded size. Stem packages are large and a network interruption easily leaves a truncated zip that fails to unpack. Verify file integrity after downloading.
与全量分轨的区别
| 接口 | 模型 | 产出 |
|---|---|---|
| 本接口 | flowmusic | 人声与伴奏为主的 zip 包 |
| 全量分轨（Suno） | suno | 鼓、贝斯、吉他、键盘等全部独立轨道 |
需要更细的乐器级拆分时，用 Suno 的全量分轨接口。
| Endpoint | Model | Output |
|---|---|---|
| This endpoint | flowmusic | A zip focused on vocals and instrumental |
| All-stems separation (Suno) | suno | Every instrument on its own track: drums, bass, guitar, keys, and so on |
For finer instrument-level separation, use Suno's all-stems endpoint.
错误码
For error response formats, status codes, and retry guidance see Error codes and rate limits; for troubleshooting steps on common errors see the FAQ.
使用示例
提交一个任务并拿到 task_id：
提交成功只表示任务已受理。真正的产物要拿返回的 task_id 去任务查询接口轮询，status 变成 completed 之后才有结果。
Submit a task and get a task_id:
A successful submission only means the task was accepted. To get the actual output, poll the task query endpoint with the returned task_id; results appear only after status becomes completed.