Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/IAHispano/Applio/llms.txt

Use this file to discover all available pages before exploring further.

The tts subcommand combines two operations into a single CLI call: it first synthesizes speech from text using Microsoft Edge TTS, saving the raw synthesized audio to --output_tts_path, and then immediately passes that audio through Applio’s RVC voice-conversion pipeline to produce a final converted output at --output_rvc_path. This means you can go from a plain text string to a fully converted voice clip — with all the quality controls available in normal inference — without any manual intermediate steps.

​
Two-stage pipeline

1

Text-to-speech synthesis

Edge TTS synthesizes the input text using the chosen voice (--tts_voice) at the specified rate (--tts_rate) and saves the raw audio to --output_tts_path.
2

RVC voice conversion

The synthesized audio is passed through the RVC pipeline using the model at --pth_path and index at --index_path. The converted audio is saved to --output_rvc_path.
--tts_text and --tts_file are both listed as required by the argument parser; provide the text directly in --tts_text and supply the path to a .txt file in --tts_file. If you only have one source, supply the same value to both or an empty string for the one you do not use — the underlying TTS script decides which to prefer at runtime.

​
Flags

​
Text input

​
--tts_text
string
required
The text string to synthesize. Enclose in quotes when passing on the command line. For long or multi-line content, use --tts_file instead.
​
--tts_file
string
required
Path to a plain-text file whose contents will be synthesized. Use this for long scripts or multi-line text.

​
Voice and rate

​
--tts_voice
string
required
Edge TTS voice short name to use for synthesis (e.g., en-US-AriaNeural, en-GB-SoniaNeural, es-ES-ElviraNeural). The full list of available voices is loaded from rvc/lib/tools/tts_voices.json; the relevant field is ShortName.
​
--tts_rate
integer
default:"0"
Speaking rate adjustment. Range: -100 (much slower) to 100 (much faster). A value of 0 uses the voice’s natural speaking rate.

​
Output paths

​
--output_tts_path
string
required
Full path where the raw Edge TTS audio will be saved before voice conversion. The file will be overwritten if it already exists inside the assets/ directory.
​
--output_rvc_path
string
required
Full path where the final RVC-converted audio will be saved.

​
Model paths

​
--pth_path
string
required
Full path to the trained RVC model file (.pth).
​
--index_path
string
required
Full path to the FAISS index file (.index) that accompanies the model.

​
Voice conversion settings

​
--pitch
integer
default:"0"
Pitch shift in semitones applied during RVC conversion. Range: -24 to 24. Useful for adapting a model trained on a different pitch register to the TTS voice’s natural pitch range.
​
--index_rate
float
default:"0.3"
Influence of the FAISS index on the conversion output. Range: 0.0–1.0. Higher values bring the output closer to the training voice; lower values reduce artifacts.
​
--volume_envelope
float
default:"1.0"
Blends the output’s volume envelope with the input’s. Range: 0.0–1.0. A value of 1.0 uses the output envelope fully.
​
--protect
float
default:"0.33"
Protects consonants and breathing sounds from conversion artifacts. Range: 0.0–0.5.
​
--f0_method
string
default:"rmvpe"
Pitch-extraction algorithm for RVC conversion. Choices: crepe, crepe-tiny, rmvpe, fcpe, hybrid[crepe+rmvpe], hybrid[crepe+fcpe], hybrid[rmvpe+fcpe], hybrid[crepe+rmvpe+fcpe]. rmvpe is recommended for most TTS voices.
​
--split_audio
boolean
default:"False"
Split the synthesized TTS audio into smaller segments before RVC conversion. Recommended for long TTS outputs. Accepts True or False.
​
--f0_autotune
boolean
default:"False"
Apply a light autotune to the RVC-converted output. Can help with TTS-to-singing use cases. Accepts True or False.
​
--f0_autotune_strength
float
default:"1.0"
Strength of the autotune snap to the chromatic grid. Range: 0.0–1.0. Only active when --f0_autotune True.
​
--proposed_pitch
boolean
default:"False"
Enable proposed pitch adjustment mode during conversion. Accepts True or False.
​
--proposed_pitch_threshold
float
default:"155.0"
Threshold frequency (Hz) for proposed pitch adjustment. Range: 100–499. Only active when --proposed_pitch True.
​
--clean_audio
boolean
default:"False"
Run noise reduction on the converted output. Recommended for cleaner TTS results. Accepts True or False.
​
--clean_strength
float
default:"0.7"
Intensity of the noise-reduction pass. Range: 0.0–1.0. Only active when --clean_audio True.
​
--export_format
string
default:"WAV"
Output file format for the final converted audio. Choices: WAV, MP3, FLAC, OGG, M4A.
​
--embedder_model
string
default:"contentvec"
Speaker-embedding model used during RVC conversion. Choices: contentvec, spin, spin-v2, chinese-hubert-base, japanese-hubert-base, korean-hubert-base, custom.
​
--embedder_model_custom
string
default:"None"
Path to a custom embedding model. Only used when --embedder_model custom.

​
Usage example

python core.py tts \
  --tts_text "Welcome to Applio voice conversion." \
  --tts_file "" \
  --tts_voice en-US-JennyNeural \
  --tts_rate 0 \
  --pth_path logs/MyModel/MyModel.pth \
  --index_path logs/MyModel/MyModel.index \
  --output_tts_path assets/tts_out.wav \
  --output_rvc_path assets/final_out.wav \
  --pitch 0 \
  --f0_method rmvpe \
  --index_rate 0.3 \
  --clean_audio True \
  --export_format WAV

​
Finding available voices

Applio loads its voice list from rvc/lib/tools/tts_voices.json. Each entry in the file represents one Edge TTS voice. The value you pass to --tts_voice must match the ShortName field exactly, including capitalization. 
# Print all available voice short names
python -c "
import json
with open('rvc/lib/tools/tts_voices.json') as f:
    voices = json.load(f)
for v in sorted({v['ShortName'] for v in voices}):
    print(v)
"
Some commonly used voices:
Short nameLanguageGender
en-US-AriaNeuralEnglish (US)Female
en-US-JennyNeuralEnglish (US)Female
en-US-GuyNeuralEnglish (US)Male
en-GB-SoniaNeuralEnglish (UK)Female
es-ES-ElviraNeuralSpanish (Spain)Female
fr-FR-DeniseNeuralFrenchFemale
de-DE-KatjaNeuralGermanFemale
ja-JP-NanamiNeuralJapaneseFemale
zh-CN-XiaoxiaoNeuralChinese (Mandarin)Female
The full Edge TTS voice catalog — including regional variants and styles — can be browsed at the Microsoft TTS voice gallery. Match the ShortName value shown there to what you pass to --tts_voice.

Build docs developers (and LLMs) love

Get started for freeTalk to us