Redirected from /stakira/openutau
OpenUtau

https://github.com/openutau/openutau
Admin
OpenUtau is an open-source singing synthesizer and editing environment for the UTAU community with...
Tokens:739,593
Snippets:3,967
Trust Score:7.8
License:MIT
Update:3 days ago
Show doc for...
Tokens:
Raw
### Install .NET SDK on Arch-based distro

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Install the .NET SDK on Arch-based distributions using pacman. Required before building OpenUtau.

```bash
sudo pacman -S dotnet-sdk
```

--------------------------------

### Install Wine Crossover via Homebrew (brew tap and brew install)

Source: https://github.com/openutau/openutau/wiki/Resamplers-and-Wavtools

Run these two commands in Terminal to install Wine Crossover after Homebrew is installed. If the brew command is not found, refer to the linked Stack Overflow thread about 'zsh command not found'.

```bash
    brew tap gcenx/wine
    brew install --cask --no-quarantine wine-crossover
```

--------------------------------

### Install .NET SDK on Ubuntu/Debian

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Install the .NET 8.0 SDK on Ubuntu, Debian, or derivatives using apt. Required before building OpenUtau.

```bash
sudo apt install dotnet-sdk-8.0
```

--------------------------------

### ops API Example Response (Renderer with ML vocoder)

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Example response from a renderer using an ML vocoder, showing ops for phonemize and synth_audio_samples. Includes optional curve parameters with ranges and scales.

```json
{
  "name": "SomeRenderer",
  "version": "1.0",
  "ops": [
    {
      "op": "phonemize",
      "inputs": {
        "note_sequence": {
          "required": true,
        }
      },
      "outputs": {
        "f0": {
          "time_unit": "ms",
          "frame_duration": 5
        },
        "phoneme_sequence": {
        }
      }
    },
    {
      "op": "synth_audio_samples",
      "inputs": {
        "f0": {
          "required": true,
          "time_unit": "ms",
          "frame_duration": 5
        },
        "phoneme_sequence": {
          "required": true
        },
        "curve:gender": {
          "required": false,
          "range": [-100, 100],
          "scale": "linear",
        },
        "curve:tension": {
          "required": false,
          "range": [-100, 100],
          "scale": "linear",
        },
        "curve:strength": {
          "required": false,
          "range": [-20, 20],
          "scale": "db",
        }
      },
      "outputs": {
        "audio_samples": {
          "channels": 1,
          "sample_rate": 44100,
          "sample_format": "int16"
        }
      }
    }
  ]
}
```

--------------------------------

### character.txt example

Source: https://github.com/openutau/openutau/wiki/Adding-support-for-a-new-synthesis-engine

Example character.txt file containing basic voicebank information such as name, image, author, voice, and web. Uses INI format with key=value pairs.

```ini
name=Zhibin Diffsinger
image=zhibin.png
author=Chisong
voice=Chisong
web=http://zhibin.club/
```

--------------------------------

### ops API Example Response (NNSVS)

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Example response from a backend using NNSVS, showing the ops list with phonemize and synth_world_features operations. Op names are free for backends to define, and input/output data structs are predefined.

```json
{
  "name": "ENUNU",
  "version": "0.5.0",
  "ops": [
    {
      "op": "phonemize", // op names are free for backends to define
      "inputs": {
        "note_sequence": {
          "required": true,
        }
      }, // input and output data structs are predefined as above
      "outputs": {
        "phoneme_sequence": {
        }
      }
    },
    {
      "op": "synth_world_features",
      "inputs": {
        "phoneme_sequence": {
          "required": true,
        }
      },
      "outputs": {
        "world_mgc": {
          "time_unit": "ms",
          "frame_duration": 5,
          "width": 60
        },
        "world_bap": {
          "time_unit": "ms",
          "frame_duration": 5,
          "width": 5
        },
        "f0": {
          "time_unit": "ms",
          "frame_duration": 5
        }
      }
    }
  ]
}
```

--------------------------------

### NNSVS synth_world_features API Example Request

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Example request to the NNSVS synth_world_features operation, passing a phoneme_sequence as defined earlier.

```json
{
  "op": "synth_world_features",
  "phoneme_sequence": [...] // as defined earlier
}
```

--------------------------------

### character.yaml example

Source: https://github.com/openutau/openutau/wiki/Adding-support-for-a-new-synthesis-engine

Example character.yaml file specifying text file encoding, portrait opacity, default phonemizer, and singer type. The singer_type is set to 'diffsinger'.

```yaml
text_file_encoding: utf-8
portrait_opacity: 0.67
default_phonemizer: OpenUtau.Core.DiffSinger.DiffSingerPhonemizer
singer_type: diffsinger
```

--------------------------------

### Build worldline with Bazelisk

Source: https://github.com/openutau/openutau/blob/master/cpp/README.md

Run from the cpp folder after installing Bazelisk. Produces a dll on Windows, dylib on macOS, or so on Linux. On Windows, omit the // in commands.

```bash
bazelisk build //worldline
```

--------------------------------

### NNSVS phonemize API Example Request

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Example request to the NNSVS phonemize operation, passing a note_sequence with a phoneme_sequence as defined earlier.

```json
{
  "op": "phonemize",
  "note_sequence": {
    "phoneme_sequence": [...] // as defined earlier
  }
}
```

--------------------------------

### ops API Example Request

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Request to the ops API to list all available operations of a backend. The response helps the frontend understand how to call the APIs to transform note sequences to audios.

```json
{
  "op": "ops"
}
```

--------------------------------

### NNSVS phonemize API Example Response

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Example response from the NNSVS phonemize operation, returning a phoneme_sequence as defined earlier.

```json
{
  "phoneme_sequence": [...] // as defined earlier
}
```

--------------------------------

### NNSVS synth_world_features API Example Response

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

Example response from the NNSVS synth_world_features operation, returning world_mgc, world_bap, and f0 arrays.

```json
{
  "world_mgc": [...],
  "world_bap": [...],
  "f0": [...]
}
```

--------------------------------

### Clone OpenUtau repository

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Clone the OpenUtau source code from GitHub using Git. Run after installing prerequisites.

```bash
git clone https://github.com/stakira/OpenUtau.git
```

--------------------------------

### Custom pronunciation dictionary YAML example

Source: https://github.com/openutau/openutau/wiki/Developing-new-phonemizers

Example of a custom dictionary in YAML 1.2 format, defining phoneme symbols and a grapheme-to-phoneme entry. Use UTF-8 encoding and ruamel.yaml for parsing/generating, not pyyaml.

```yaml
%YAML 1.2
---
symbols:
  - {symbol: aa, type: vowel}
  - {symbol: ae, type: vowel}
  - {symbol: ah, type: vowel}
  - {symbol: ao, type: vowel}
  - {symbol: aw, type: vowel}
  - {symbol: ay, type: vowel}
  - {symbol: b, type: stop}
  - {symbol: ch, type: affricate}
  - {symbol: d, type: stop}
  - {symbol: dh, type: fricative}
  - {symbol: eh, type: vowel}
  - {symbol: er, type: vowel}
  - {symbol: ey, type: vowel}
  - {symbol: f, type: fricative}
  - {symbol: g, type: stop}
  - {symbol: hh, type: aspirate}
  - {symbol: ih, type: vowel}
  - {symbol: iy, type: vowel}
  - {symbol: jh, type: affricate}
  - {symbol: k, type: stop}
  - {symbol: l, type: liquid}
  - {symbol: m, type: nasal}
  - {symbol: n, type: nasal}
  - {symbol: ng, type: nasal}
  - {symbol: ow, type: vowel}
  - {symbol: oy, type: vowel}
  - {symbol: p, type: stop}
  - {symbol: r, type: liquid}
  - {symbol: s, type: fricative}
  - {symbol: sh, type: fricative}
  - {symbol: t, type: stop}
  - {symbol: th, type: fricative}
  - {symbol: uh, type: vowel}
  - {symbol: uw, type: vowel}
  - {symbol: v, type: fricative}
  - {symbol: w, type: semivowel}
  - {symbol: y, type: semivowel}
  - {symbol: z, type: fricative}
  - {symbol: zh, type: fricative}
entries:
  - grapheme: openutau
    phonemes: [ow, p, eh, n, w, uw, t, ah, w, uw]

```

--------------------------------

### Find Wine path with which wine

Source: https://github.com/openutau/openutau/wiki/Resamplers-and-Wavtools

Run this command in Terminal to locate the Wine installation path. The path differs between Intel and Silicon Macs.

```bash
    which wine
```

--------------------------------

### Full Moresampler Resampler Manifest (YAML)

Source: https://github.com/openutau/openutau/wiki/Resamplers-and-Wavtools

Example manifest for moresampler.exe, named moresampler.yaml and stored in the same folder. Lists supported expressions with type, range, default, and flag. Add `expression_filter: true` to disable unlisted flags.

```yaml
expressions:
  gen:
    name: Gender Factor
    abbr: gen
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: g
  mbre:
    name: Breathiness (Moresampler)
    abbr: mbre
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: Mb
  tens:
    name: Tension
    abbr: tens
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: Mt
  pit:
    name: Pitch deviation (flag)
    abbr: pit
    type: Numerical
    min: -1200
    max: 1200
    default_value: 0
    is_flag: true
    flag: t
  pkcp:
    name: Peak Compressor
    abbr: pkcp
    type: Numerical
    min: 0
    max: 100
    default_value: 86
    is_flag: true
    flag: P
  amp:
    name: Amplitude Modulation
    abbr: amp
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: A
  cons:
    name: Unvoiced Consonant Gain
    abbr: cons
    type: Numerical
    min: -20
    max: 100
    default_value: 0
    is_flag: true
    flag: b
  fstr:
    name: Force Stretch
    abbr: fstr
    type: Options
    min: 0
    max: 1
    default_value: 0
    is_flag: true
    options:
    - ''
    - e
    - Me
  opn:
    name: Openness
    abbr: opn
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: Mo
  res:
    name: Resonance
    abbr: res
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: Mr
  dry:
    name: Dryness
    abbr: dry
    type: Numerical
    min: -100
    max: 100
    default_value: 0
    is_flag: true
    flag: Md
  cors:
    name: Coarseness
    abbr: cors
    type: Numerical
    min: 0
    max: 100
    default_value: 0
    is_flag: true
    flag: MC
  grwl:
    name: Growl
    abbr: grwl
    type: Numerical
    min: 0
    max: 100
    default_value: 0
    is_flag: true
    flag: MG
  dist:
    name: Distortion
    abbr: dist
    type: Numerical
    min: 0
    max: 100
    default_value: 0
    is_flag: true
    flag: MD
  stbl:
    name: Stabilization
    abbr: stbl
    type: Numerical
    min: 0
    max: 10
    default_value: 0
    is_flag: true
    flag: Ms
  mint:
    name: Model Interpolation
    abbr: mint
    type: Numerical
    min: 0
    max: 100
    default_value: 100
    is_flag: true
    flag: Mm

```

--------------------------------

### YAML custom pronunciation dictionary example

Source: https://github.com/openutau/openutau/wiki/Compiling-G2p-Models

Example custom dictionary for G2pDictionary. Uses YAML 1.2 syntax and must be UTF-8 encoded regardless of the voicebank encoding. Defines symbols and a grapheme-to-phoneme entry for 'openutau'.

```yaml
%YAML 1.2
---
symbols:
  - {symbol: aa, type: vowel}
  - {symbol: ae, type: vowel}
  - {symbol: ah, type: vowel}
  - {symbol: ao, type: vowel}
  - {symbol: aw, type: vowel}
  - {symbol: ay, type: vowel}
  - {symbol: b, type: stop}
  - {symbol: ch, type: affricate}
  - {symbol: d, type: stop}
  - {symbol: dh, type: fricative}
  - {symbol: eh, type: vowel}
  - {symbol: er, type: vowel}
  - {symbol: ey, type: vowel}
  - {symbol: f, type: fricative}
  - {symbol: g, type: stop}
  - {symbol: hh, type: aspirate}
  - {symbol: ih, type: vowel}
  - {symbol: iy, type: vowel}
  - {symbol: jh, type: affricate}
  - {symbol: k, type: stop}
  - {symbol: l, type: liquid}
  - {symbol: m, type: nasal}
  - {symbol: n, type: nasal}
  - {symbol: ng, type: nasal}
  - {symbol: ow, type: vowel}
  - {symbol: oy, type: vowel}
  - {symbol: p, type: stop}
  - {symbol: r, type: liquid}
  - {symbol: s, type: fricative}
  - {symbol: sh, type: fricative}
  - {symbol: t, type: stop}
  - {symbol: th, type: fricative}
  - {symbol: uh, type: vowel}
  - {symbol: uw, type: vowel}
  - {symbol: v, type: fricative}
  - {symbol: w, type: semivowel}
  - {symbol: y, type: semivowel}
  - {symbol: z, type: fricative}
  - {symbol: zh, type: fricative}
entries:
  - grapheme: openutau
    phonemes: [ow, p, eh, n, w, uw, t, ah, w, uw]
```

--------------------------------

### Full character.yaml example with subbanks

Source: https://github.com/openutau/openutau/wiki/tech-note-‐-character.yaml

Complete example based on 闇音レンリ・連続音Ver1.5, a multi-pitch, multi-color bank with 3 voice colors (Normal, Whisper, Clear) and 4 tone ranges (A3, D4, G4, C5), plus extras ↑ and Edge. The subbanks list defines 14 subbanks with suffixes and tone ranges. Every field is optional.

```yaml
name: 闇音レンリ・連続音Ver1.5
localized_names:
  en-US: Renri Yamine VCV Ver1.5
singer_type: utau
text_file_encoding: shift_jis
image: renri.bmp
portrait: portrait.webp
portrait_opacity: 0.67
portrait_height: 0
author: ゆずり
voice: ゆずり
version: 1.5
web: https://renrivoice.wixsite.com/renri-voice/utau
default_phonemizer: OpenUtau.Plugin.Builtin.JapaneseVCVPhonemizer
subbanks:
- color: ''
  prefix: ''
  suffix: C5
  tone_ranges:
  - C5-B7
- color: ''
  prefix: ''
  suffix: G4
  tone_ranges:
  - G4-B4
- color: ''
  prefix: ''
  suffix: D4
  tone_ranges:
  - D4-F#4
- color: ''
  prefix: ''
  suffix: A3
  tone_ranges:
  - C1-C#4
- color: ↑
  prefix: ''
  suffix: ↑
  tone_ranges:
  - C1-B7
- color: Clear
  prefix: ''
  suffix: CC5
  tone_ranges:
  - C5-B7
- color: Clear
  prefix: ''
  suffix: CG4
  tone_ranges:
  - G4-B4
- color: Clear
  prefix: ''
  suffix: CD4
  tone_ranges:
  - D4-F#4
- color: Clear
  prefix: ''
  suffix: CA3
  tone_ranges:
  - C1-C#4
- color: Whisper
  prefix: ''
  suffix: WC5
  tone_ranges:
  - C5-B7
- color: Whisper
  prefix: ''
  suffix: WG4
  tone_ranges:
  - G4-B4
- color: Whisper
  prefix: ''
  suffix: WD4
  tone_ranges:
  - D4-F#4
- color: Whisper
  prefix: ''
  suffix: WA3
  tone_ranges:
  - C1-C#4
- color: Edge
  prefix: ''
  suffix: "'"
  tone_ranges:
  - C1-B7
```

--------------------------------

### phonemize

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

An example API operation that converts a note sequence into a phoneme sequence. The request and response are msgpack objects with the op name and required data structs.

```APIDOC
## phonemize

### Description
Converts a note sequence into a phoneme sequence. This is an example operation defined by a backend (e.g., NNSVS).

### Method
POST (implied by request/response format)

### Endpoint
Not specified (backend-defined)

### Parameters
#### Request Body
- **op** (string) - Required - Operation name, must be "phonemize"
- **note_sequence** (object) - Required - Contains the note sequence data, with a `phoneme_sequence` field (as defined earlier)

### Request Example
{
  "op": "phonemize",
  "note_sequence": {
    "phoneme_sequence": []
  }
}

### Response
#### Success Response
- **phoneme_sequence** (array) - The resulting phoneme sequence

#### Response Example
{
  "phoneme_sequence": []
}
```

--------------------------------

### Expression Definition Example (dyn)

Source: https://github.com/openutau/openutau/wiki/USTX-file-format

Shows the YAML structure for defining a vocal parameter expression under the 'expressions' section. This example defines a curve-type expression named 'dynamics (curve)' with abbreviation 'dyn', a value range from -240 to 120, and a default value of 0.

```yaml
dyn:
  name: dynamics (curve)
  abbr: dyn
  type: Curve
  min: -240
  max: 120
  default_value: 0

```

--------------------------------

### ops

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

The ops API describes all available operations of a backend. It helps the frontend understand how to call the APIs to transform note sequences to audios. This is the only required API for a backend and can also be used to check if the backend is started or still alive.

```APIDOC
## ops

### Description
Describes all available operations of a backend. Helps the frontend understand how to call the APIs to transform note sequences to audios. This is the only required API for a backend. Can also be used to check if the backend is started or still alive; if the backend does not return in 1 second, it is considered offline.

### Method
POST (implied by request/response format)

### Endpoint
Not specified (backend-defined)

### Parameters
#### Request Body
- **op** (string) - Required - Must be "ops"

### Request Example
{
  "op": "ops"
}

### Response
#### Success Response
- **name** (string) - Backend name
- **version** (string) - Backend version
- **ops** (array) - List of available operations, each with:
  - **op** (string) - Operation name (free for backends to define)
  - **inputs** (object) - Input data structs, each with optional `required` boolean
  - **outputs** (object) - Output data structs, each with optional metadata like `time_unit`, `frame_duration`, `width`, `channels`, `sample_rate`, `sample_format`

#### Response Example
{
  "name": "ENUNU",
  "version": "0.5.0",
  "ops": [
    {
      "op": "phonemize",
      "inputs": {
        "note_sequence": {
          "required": true
        }
      },
      "outputs": {
        "phoneme_sequence": {}
      }
    },
    {
      "op": "synth_world_features",
      "inputs": {
        "phoneme_sequence": {
          "required": true
        }
      },
      "outputs": {
        "world_mgc": {
          "time_unit": "ms",
          "frame_duration": 5,
          "width": 60
        },
        "world_bap": {
          "time_unit": "ms",
          "frame_duration": 5,
          "width": 5
        },
        "f0": {
          "time_unit": "ms",
          "frame_duration": 5
        }
      }
    }
  ]
}
```

--------------------------------

### Define fallbacks for missing phonemes

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Maps missing phonemes or aliases to existing ones. Only 1:1 (one-to-one) mappings are supported. The example shows both inline and block formats.

```yaml
fallbacks:
# 1:1 (one-to-one mappings)
  - {from: a, to: A}
  - {from: ts, to: ch}

# format can be like this
    - from: b@
    to: bV
    - from: b
    to: bh
```

--------------------------------

### synth_world_features

Source: https://github.com/openutau/openutau/wiki/[PROPOSAL]-svs.io-‐-singing-voice-synthesis-backend-API

An example API operation that synthesizes WORLD features (mgc, bap, f0) from a phoneme sequence. The request and response are msgpack objects with the op name and required data structs.

```APIDOC
## synth_world_features

### Description
Synthesizes WORLD features (mgc, bap, f0) from a phoneme sequence. This is an example operation defined by a backend (e.g., NNSVS).

### Method
POST (implied by request/response format)

### Endpoint
Not specified (backend-defined)

### Parameters
#### Request Body
- **op** (string) - Required - Operation name, must be "synth_world_features"
- **phoneme_sequence** (array) - Required - The phoneme sequence input

### Request Example
{
  "op": "synth_world_features",
  "phoneme_sequence": []
}

### Response
#### Success Response
- **world_mgc** (array) - Mel-generalized cepstrum features
- **world_bap** (array) - Band aperiodicity features
- **f0** (array) - Fundamental frequency values

#### Response Example
{
  "world_mgc": [],
  "world_bap": [],
  "f0": []
}
```

--------------------------------

### config.yaml style_format and styles for expression control

Source: https://github.com/openutau/openutau/wiki/ENUNU-NNSVS-Support

Define style_format and styles in config.yaml to control expressions other than voice colour. The example defines two flag keys (p9 and p16) with their formats and index keys, then lists each expression style with its name, type, min/max, default value, and flag. Only numeric expression types are currently supported.

```yaml
extensions:
# Other settings....
  style_format:
    p9: #Flag key to write to UST
      format: "{0}{1}{2}{3}" #Flag format Please enclose in "
      index: [ldst, loud, nrml, soft] #Key Set for Expression
    p16: 
      format: "s{0}s{1}r{2}b{3}p{4}"
      index: [Std, swt, rock, brth, pop]
  styles:
    ldst: #Key to Expression
      name: Loudest #Expression name
      type: Numerical #Expression control type: Currently, only numeric values are supported.
      min: 0
      max: 3
      default_value: 0
      flag: p9 #Flag key to write to UST
      
    loud:
      name: Loud 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p9 
      
    nrml:
      name: Normal 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p9 
      
    soft:
      name: Soft 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p9 
      
    Std:
      name: Standard 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p16 
      
    swt:
      name: Sweet 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p16 
      
    rock:
      name: Rock 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p16 
      
    brth:
      name: Breathy 
      type: Numerical 
      min: 0
      max: 6
      default_value: 0
      flag: p16 
      
    pop:
      name: Pop 
      type: Numerical 
      min: 0
      max: 3
      default_value: 0
      flag: p16 

```

--------------------------------

### Symbols configuration in njokis.yaml

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Defines custom symbols and their types for the phonemizer. Encapsulate the whole symbol with single quotes if it starts with a special symbol.

```yaml
symbols:
    - {symbol: ea, type: vowel}
    - {symbol: ix, type: vowel}
    - {symbol: dd, type: tap}
    - {symbol: j, type: fricative}
    - {symbol: '@r', type: vowel}
```

--------------------------------

### EN ARPA+ Symbols Configuration

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Defines custom symbols and their types for the ARPAsing+ phonemizer. Use 'tail' for trailing tail notes and wrap symbols starting with special characters in single quotes.

```yaml
symbols:
  - {symbol: ea, type: vowel}
  - {symbol: ix, type: vowel}
  - {symbol: dd, type: tap}
  - {symbol: j, type: fricative}
  - {symbol: '@r', type: vowel}
```

--------------------------------

### init (request)

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

Sent exactly once, immediately after connecting, with the full USTX project document as the baseline for all later incremental updates. The response data contains the API version the plugin implements.

```APIDOC
## init (request)

### Description
Sent exactly once, immediately after connecting, with the full USTX project document as the baseline for all later incremental updates. The response data contains the API version the plugin implements.

### Message Type
Request

### Payload
{
  "ustx": "<full USTX project document>"
}

### Response Data
{
  "apiVersion": "1.2"
}

### Notes
- The `ustx` field is OpenUtau's native USTX document, which is YAML — byte-identical to what OpenUtau writes into a `.ustx` file. A plugin may persist it or re-parse it with any YAML reader. It is deliberately **not** a JSON projection of the in-memory project object, which would be lossy.
- OpenUtau is the sole owner of the project: the baseline travels outward only and is never echoed back.
```

--------------------------------

### Restore and build OpenUtau with dotnet

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Restore NuGet packages and build the OpenUtau project. The executable files will be placed in `OpenUtau/OpenUtau/bin/Debug/net6.0`.

```bash
dotnet restore OpenUtau

dotnet build OpenUtau
```

--------------------------------

### Navigate to OpenUtau directory

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Change to the OpenUtau directory before building from the command line.

```bash
cd OpenUtau
```

--------------------------------

### getAudio (request)

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

Requests the rendered audio for a hash advertised in updatePartLayout. The response is a data-plane frame containing raw float32 PCM audio.

```APIDOC
## getAudio (request)

### Description
Requests the rendered audio for a hash advertised in `updatePartLayout`. The response is a data-plane frame, not a JSON envelope.

### Method
request

### Endpoint
getAudio

### Parameters
#### Request Body
- **hash** (string) - Required - The hash of the audio to pull.

### Request Example
```json
{ "hash": "13507256038857166760" }
```

### Response
#### Success Response
Raw audio frame: `audio <hash> <byteCount>\n<raw bytes>`

Payload encoding: raw `float32` PCM, 44.1 kHz, stereo, interleaved, little-endian. No compression, no base64. Mono mixes are upmixed to stereo by OpenUtau before transmission.

### Notes
- The plugin SHOULD pull missing hashes sequentially; this version allows one outstanding `getAudio` per connection.
- The plugin is the requester by design; pulling rather than being pushed to gives the plugin backpressure over a slow DAW track.
```

--------------------------------

### Switch back to original branch after testing

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Use after testing to return to your own branch, then pop the stashed changes to restore uncommitted work.

```bash
git checkout {branch}
```

```bash
git stash pop
```

--------------------------------

### Clean Bazel build cache

Source: https://github.com/openutau/openutau/blob/master/cpp/README.md

Cleans up the build cache. Use --expunge to also remove dependencies.

```bash
bazelisk clean
```

--------------------------------

### Clean Bazel build cache and dependencies

Source: https://github.com/openutau/openutau/blob/master/cpp/README.md

Cleans up the build cache and dependencies. Use if you need a full reset.

```bash
bazelisk clean --expunge
```

--------------------------------

### Classic packaging format folder structure

Source: https://github.com/openutau/openutau/wiki/Adding-support-for-a-new-synthesis-engine

Shows the folder structure of the classic packaging format used by UTAU, ENUNU, and DiffSinger voicebanks. The structure includes character.txt, character.yaml, and other voice data files.

```text
your_singer
├─ character.txt
├─ character.yaml
├─ ...the other files and folders containing the voice data.
```

--------------------------------

### playbackStarted (notification)

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

Sent on the DAW transport's play rising edge. OpenUtau flushes all pending debounced updates before playback begins.

```APIDOC
## playbackStarted (notification)

### Description
Sent on the DAW transport's play rising edge. OpenUtau flushes all pending debounced updates (§7) before playback begins, so the plugin plays against current state.

### Method
notification

### Endpoint
playbackStarted

### Request Example
```json
{}
```
```

--------------------------------

### updatePartLayout (request)

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

Provides the current part layout on the shared timeline, including audio hashes for deduplication. The response data lists the hashes the receiver still needs.

```APIDOC
## updatePartLayout (request)

### Description
Provides the current part layout on the shared timeline, including audio hashes for deduplication. The response data lists the hashes the receiver still needs.

### Message Type
Request

### Payload
{
  "parts": [
    { "trackNo": 0, "startMs": 1200.0, "endMs": 8400.0,
      "audioHash": "13507256038857166760" }
  ]
}

### Payload Fields
- **trackNo** (integer) - Zero-based index into the last `updateTracks` list
- **startMs** (number) - Part start on the shared timeline, milliseconds
- **endMs** (number) - Part end, same coordinate system
- **audioHash** (string) - XXH64 (decimal) of that part's rendered audio

### Response Data
{
  "missingAudios": ["13507256038857166760", "…"]
}

### Notes
- The receiver deduplicates against audio it already holds. It **SHOULD** additionally verify a received frame's `length` against its own hash computation as a cheap integrity cross-check.
- An empty array in `missingAudios` means everything is cached.
```

--------------------------------

### Fetch and checkout a pull request branch

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Replace {author} and {branch} with the pull request author and branch name (e.g., git fetch https://github.com/lennyservant/OpenUtau improve-classic-pitch-timing; git checkout FETCH_HEAD). After testing, switch back with git checkout {branch} and restore changes with git stash pop.

```bash
git fetch https://github.com/{author}/OpenUtau {branch}; git checkout FETCH_HEAD
```

--------------------------------

### bpm (notification) — since 1.1

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

Sends the DAW project's tempo when it changes. OpenUtau uses this only as a guard to warn the user about bar misalignment.

```APIDOC
## bpm (notification) — since 1.1

### Description
The DAW project's tempo, sent when it changes. OpenUtau uses this only as a guard — warning the user, once per distinct value outside a **±0.5 BPM** tolerance, that bars will misalign. It never retempo-maps the project (no tempo-map sync in this version).

### Method
notification

### Endpoint
bpm

### Parameters
#### Request Body
- **bpm** (number) - Required - The DAW project's tempo.

### Request Example
```json
{ "bpm": 137.5 }
```
```

--------------------------------

### EN ARPA+ Timings Configuration

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Sets phoneme lengths in seconds for specific symbols or aliases.

```yaml
timings:
  - {symbol: dh, value: 1.2}
  - {symbol: vf, value: 2.8}
  - {symbol: st, value: 3.5}
  - {symbol: a r, value: 0.5}
  - {symbol: e r, value: 0.5}
  - {symbol: i r, value: 0.5}
```

--------------------------------

### Silicon Mac Wine path (Homebrew)

Source: https://github.com/openutau/openutau/wiki/Resamplers-and-Wavtools

For Silicon (Apple Silicon) users, the Wine path is inside Homebrew.

```bash
    /opt/homebrew/bin/wine
```

--------------------------------

### Intel Mac Wine path (wine32on64)

Source: https://github.com/openutau/openutau/wiki/Resamplers-and-Wavtools

For Intel users, the Wine path will contain either 32 or 64, as shown here.

```bash
    /usr/local/bin/wine32on64
```

--------------------------------

### Fix Resources.Designer.cs namespace before adding G2p data

Source: https://github.com/openutau/openutau/wiki/Compiling-G2p-Models

After dragging the G2p zip into Resources.resx, Resources.Designer.cs incorrectly points to 'G2p.Data.Resources'. Replace it with 'OpenUtau.Core.G2p.Data.Resources' on line 42 so the G2p works properly.

```csharp
if (object.ReferenceEquals(resourceMan, null)) {
                    global::System.Resources.ResourceManager temp = new global::System.Resources.ResourceManager("G2p.Data.Resources", typeof(Resources).Assembly);
                    resourceMan = temp;
                }
```

```csharp
if (object.ReferenceEquals(resourceMan, null)) {
                    global::System.Resources.ResourceManager temp = new global::System.Resources.ResourceManager("OpenUtau.Core.G2p.Data.Resources", typeof(Resources).Assembly);
                    resourceMan = temp;
                }
```

--------------------------------

### EN ARPA+ Fallbacks Configuration

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Maps missing phonemes or aliases to existing ones. Only 1:1 mappings are supported.

```yaml
fallbacks:
# 1:1 (one-to-one mappings)
  - {from: ax, to: ah}
  - {from: ch oy, to: ch ow}

# format can be like this
    - from: b@
    to: bV
    - from: b
    to: bh
```

--------------------------------

### SimpleEnunuServer-0.5.0 config.yaml extensions

Source: https://github.com/openutau/openutau/wiki/ENUNU-NNSVS-Support

Configuration for SimpleEnunuServer-0.5.0. Place in config.yaml. Only the timing_editor_2 extension is specified for this version.

```yaml
extensions:
    timing_editor_2: "%v/timing_auto_correct/enunu_timing_auto_correct.py"

```

--------------------------------

### updateProjectInfo (notification) — since 1.1

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

Provides project name and saved state for display in plugin UI. Carries no state the mixer or renderer needs.

```APIDOC
## updateProjectInfo (notification) — since 1.1

### Description
Provides project name and saved state for display in plugin UI. Carries no state the mixer or renderer needs.

### Message Type
Notification

### Payload
{ "name": "my song", "saved": true }

### Payload Fields
- **name** (string) - The project file's stem; an unsaved project reports an empty `name`.
- **saved** (bool) - Whether the project is saved.

### Notes
- What a plugin's UI may show about the project. `name` is the project file's stem; an unsaved project reports `saved` `false` and an empty `name`.
- Carries no state the mixer or renderer needs; a plugin may ignore it.
```

--------------------------------

### playhead (notification) — since 1.1

Source: https://github.com/openutau/openutau/blob/master/OpenUtau.Core/DawIntegration/API.md

One-way notification towards OpenUtau with the DAW transport position. The received position overwrites OpenUtau's playhead, converted to ticks on the project's own time axis.

```APIDOC
## playhead (notification) — since 1.1

### Description
One-way notification towards OpenUtau with the DAW transport position. The received position overwrites OpenUtau's playhead, converted to ticks on the project's own time axis. There is no reverse direction, reply, or acknowledgement.

### Method
notification

### Endpoint
playhead

### Parameters
#### Request Body
- **positionMs** (number) - Required - DAW transport position, absolute milliseconds on the shared timeline — the same coordinate system as `updatePartLayout`'s `startMs`.
- **playing** (bool) - Required - DAW transport state.

### Request Example
```json
{ "positionMs": 12500.0, "playing": false }
```

### Notes
- Moves smaller than **5 ticks** at the destination are ignored as jitter.
- Pacing is the sender's choice; receivers MUST tolerate any pacing. The reference plugin sends state changes immediately, every 100 ms while playing, and only when a parked playhead moves more than 50 ms.
```

--------------------------------

### EN ARPA+ Entries Configuration

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Adds custom dictionary entries to override word pronunciations. Supports any Unicode characters.

```yaml
entries:
  - {grapheme: hello, phonemes: [hh, ax, l, ow]}
  - {grapheme: 안녕하세요, phonemes: [aa, n, y, ao, ng, hh, ah, s, eh, y ao]}
# format can be like this
  - grapheme: stars
    phonemes: s, t, aa, r, z
  - grapheme: drops
    phonemes: dr, aa, p, s
```

--------------------------------

### Portrait settings (portrait, portrait_opacity, portrait_height)

Source: https://github.com/openutau/openutau/wiki/tech-note-‐-character.yaml

Configures the portrait image displayed on the Piano Roll. A relative path like 'images\portrait.png' works. Unless portrait_height is defined, the image is displayed in its original size unless taller than 800px, in which case it is resized to 800 tall. Setting portrait_height to 0 uses default values so smaller images don't scale up. After editing while OpenUtau is open, use Singers dialog -> '...' -> 'Refresh' to reload.

```yaml
portrait: portrait.png
portrait_opacity: 0.67
portrait_height: 0
```

--------------------------------

### Basic character info fields

Source: https://github.com/openutau/openutau/wiki/tech-note-‐-character.yaml

Provides basic character information similar to character.txt. If these fields are not present, the info in character.txt will be used.

```yaml
name: 闇音レンリ・連続音Ver1.5
singer_type: utau
image: renri.bmp
author: ゆずり
voice: ゆずり
version: 1.5
web: https://renrivoice.wixsite.com/renri-voice/utau
```

--------------------------------

### Stash uncommitted changes before testing a pull request

Source: https://github.com/openutau/openutau/wiki/Compiling-from-source

Run in the OpenUtau development folder to save any uncommitted changes before fetching and testing a pull request.

```bash
git stash
```

--------------------------------

### EN ARPA+ Replacements Configuration

Source: https://github.com/openutau/openutau/wiki/Phonemizers

Replaces phonemes or sequences with other phonemes. Supports 1:1, 1:M, M:1, and M:M mappings.

```yaml
replacements:
# 1:1 (one-to-one mappings)
  - {from: ax, to: ah}
  - {from: cl, to: q}
# 1:M (one-to-many mappings) (splitting)
  - {from: dr, to: [d, r]}
  - {from: tr, to: [t, sh, r]}
  - {from: aw, to: [aa, w]}
# M:1 (many-to-one mappings) (merging)
  - {from: [ih, ng], to: ing}
  - {from: [eh, ax, m], to: eam}
# M:M (many-to-many mappings)
  - {from: [ih, ng], to: [ix, ng]}
  - {from: [ay, l], to: [ay, ax, el]}

# format can be like this
    - from: [ae, n]
    to: [eh, ax, n]
    - from: b
    to: bh
```

--------------------------------

### ENUNU-0.4.0/SimpleEnunuServer-0.2.0 config.yaml extensions

Source: https://github.com/openutau/openutau/wiki/ENUNU-NNSVS-Support

Configuration for ENUNU-0.4.0 and SimpleEnunuServer-0.2.0+0.local.9.1. Place in config.yaml. Delete the timing_editor line to disable timing correction. Turning off wav_synthesizer allows pitch editing but results in very low quality.

```yaml
extensions:
    timing_editor: "%v/timing_auto_correct/enunu_timing_auto_correct.py"
    wav_synthesizer: synthe
    acoustic_calculator: nnsvs
    timing_calculator: nnsvs
    ust_converter: built-in

```

--------------------------------

### Trust OpenUtau app on macOS with xattr

Source: https://github.com/openutau/openutau/wiki/FAQ

Run this command in Terminal to remove macOS quarantine attributes so OpenUtau can be opened. Use the same command on the downloaded DMG if it cannot be opened.

```bash
xattr -rc /Applications/OpenUtau.app
```

```bash
xattr -rc ~/Downloads/OpenUtau-osx-x64.dmg
```

--------------------------------

### C# G2pPack field order before prioritizing dict.text

Source: https://github.com/openutau/openutau/wiki/Compiling-G2p-Models

Original field and assignment order in the G2pPack class before editing. The source shows this as the 'from' state to be changed so that dict.text is prioritized over g2p.onnx.

```csharp
private static object lockObj = new object();
private static Dictionary<string, int> graphemeIndexes;
private static IG2p dict;
private static InferenceSession session;
private static Dictionary<string, string[]> predCache = new Dictionary<string, string[]>();

// public class name of the g2p 
        GraphemeIndexes = graphemeIndexes;
        Phonemes = phonemes;
        Dict = dict;
        Session = session;
        PredCache = predCache;
```
Tokens:
Raw