Command-Line Interface

Applio provides a command-line interface through core.py that allows you to perform all major operations without the Gradio web UI. This is useful for automation, headless servers, cloud environments, and advanced scripting.

Usage

python core.py <command> [options]

Global Options

Option	Description
-h, --help	Show help message and exit

Commands

infer

Convert a single audio file using a voice model.

python core.py infer -i <input> -o <output> -m <model> [options]

Option	Alias	Description	Default
--input-audio-path	-i	Path to the input audio file	Required
--output-audio-path	-o	Path to save the converted audio	Required
--model-path	-m	Path to the model .pth file	Required
--index-path	-id	Path to the .index file	None
--pitch	-p	Pitch adjustment in semitones	0
--filter-radius	-r	Filter radius for artifact reduction	3
--search-feature-ratio	-s	Index search feature ratio (0.0 - 1.0)	0.6
--volume-envelope	-v	Volume envelope blending ratio	1
--protect-voiceless-consonants	-c	Protection amount for voiceless consonants	0.5
--hop-length	-hl	Hop length for pitch tracking	128
--f0-method	-f	Pitch extraction method (rmvpe, crepe, crepe-tiny, fcpe, hybrid)	rmvpe
--f0-file	-ff	Path to external F0 curve file	None
--embedder-model	-e	Embedder model name	contentvec
--export-format	-ef	Output audio format (wav, mp3, flac, ogg, m4a)	wav
--upscale	-u	Enable audio upscaling	False
--clean	-cl	Enable noise reduction cleaning	False
--clean-strength	-cs	Noise reduction strength	0
--autotune	-at	Enable autotune	False
--autotune-strength	-as	Autotune strength	1
--split-audio	-sa	Enable audio splitting for long files	False
--reverb	-rv	Reverb amount (0-100)	0
--pitch-shift	-ps	Post-processing pitch shift (semitones)	0
--limiter	-l	Limiter threshold (0-100)	0
--gain	-g	Gain adjustment (dB)	0
--distortion	-d	Distortion amount (0-100)	0
--chorus	-ch	Chorus amount (0-100)	0
--bitcrush	-bc	Bitcrush amount (0-100)	0
--clipping	-clp	Clipping threshold (0-100)	0
--compressor	-cmp	Compressor amount (0-100)	0
--delay	-dl	Delay amount (0-100)	0
--formant-shift	-fs	Formant shift quefrency	0
--formant-timbre	-ft	Formant shift timbre	1
--formant-q	-fq	Formant shift Q factor	5

batch_infer

Convert multiple audio files in a directory.

python core.py batch_infer -i <input_folder> -o <output_folder> -m <model> [options]

Option	Alias	Description	Default
--input-folder	-i	Path to folder containing input audio files	Required
--output-folder	-o	Path to folder for output files	Required
--model-path	-m	Path to the model .pth file	Required
All inference options		Same as infer command	Same as infer

tts

Generate speech from text using Edge TTS, optionally passing through voice conversion.

python core.py tts -t <text> -v <voice> [options]

Option	Alias	Description	Default
--text	-t	Text to synthesize	Required
--tts-voice	-v	TTS voice name (from edge-tts)	Required
--tts-rate	-tr	TTS speech rate adjustment	+0%
--model-path	-m	Path to voice model .pth for VC after TTS	None
--index-path	-id	Path to .index file for VC	None
--pitch	-p	Pitch adjustment after TTS	0
--f0-method	-f	Pitch extraction method	rmvpe
--embedder-model	-e	Embedder model	contentvec
All post-processing options		Same as infer	Same as infer

preprocess

Preprocess a dataset for training.

python core.py preprocess -n <model_name> -sr <sample_rate> [options]

Option	Alias	Description	Default
--model-name	-n	Name of the model/dataset	Required
--sample-rate	-sr	Target sample rate (32k, 40k, 48k)	40k
--cpu	-cpu	Process on CPU instead of GPU	False

extract

Extract features (F0 and embeddings) from a preprocessed dataset.

python core.py extract -n <model_name> -sr <sample_rate> -f <f0_method> -e <embedder>

Option	Alias	Description	Default
--model-name	-n	Name of the model	Required
--sample-rate	-sr	Sample rate (32k, 40k, 48k)	40k
--f0-method	-f	Pitch extraction method	rmvpe
--embedder-model	-e	Embedder model	contentvec
--cpu	-cpu	Process on CPU	False

train

Train a voice model.

python core.py train -n <model_name> -sr <sample_rate> -f <f0_method> -e <embedder> [options]

Option	Alias	Description	Default
--model-name	-n	Name of the model	Required
--sample-rate	-sr	Sample rate (32k, 40k, 48k)	40k
--f0-method	-f	Pitch extraction method	rmvpe
--embedder-model	-e	Embedder model	contentvec
--save-every-epoch	-s	Save checkpoint every N epochs	10
--total-epoch	-te	Total number of training epochs	200
--batch-size	-b	Training batch size	6
--pretrained-g	-pg	Path to pre-trained generator	None
--pretrained-d	-pd	Path to pre-trained discriminator	None
--vocoder	-v	Vocoder type (hifigan, nsf-hifigan, refinegan)	hifigan
--index	-i	Train index after training	True

index

Generate a FAISS or KMeans index from extracted features.

python core.py index -n <model_name>

Option	Alias	Description	Default
--model-name	-n	Name of the trained model	Required

model_information

Display metadata about a trained model.

python core.py model_information -m <model_path>

Option	Alias	Description	Default
--model-path	-m	Path to the .pth model file	Required

Output includes: version, sample rate, whether F0 was used, embedder name, epoch count, step count.

model_blender

Blend two model weights into a new model using linear interpolation.

python core.py model_blender -m1 <model_a> -m2 <model_b> -o <output> -r <ratio>

Option	Alias	Description	Default
--model-a	-m1	Path to first model .pth	Required
--model-b	-m2	Path to second model .pth	Required
--output	-o	Path for the blended output .pth	Required
--ratio	-r	Blend ratio (0.0 = all model A, 1.0 = all model B)	0.5

tensorboard

Launch TensorBoard for monitoring training progress.

python core.py tensorboard

download

Download a model from a URL (HuggingFace, Google Drive, or direct link).

python core.py download -u <url>

Option	Alias	Description	Default
--url	-u	Download URL for the model	Required

prerequisites

Download and install required prerequisites (pretrained models, predictors, embedders).

python core.py prerequisites

audio_analyzer

Analyze an audio file and display spectral information.

python core.py audio_analyzer -i <input_audio>

Option	Alias	Description	Default
--input-audio-path	-i	Path to the audio file to analyze	Required

Examples

Basic inference with a model:

python core.py infer -i my_speech.wav -o converted.wav -m logs/MyModel/G_1234.pth -id logs/MyModel/MyModel.index

Batch inference with pitch adjustment:

python core.py batch_infer -i ./input_folder -o ./output_folder -m logs/Singer/G_5678.pth -p -2 -f crepe

Full training pipeline:

python core.py preprocess -n MyVoice -sr 40k
python core.py extract -n MyVoice -sr 40k -f rmvpe -e contentvec
python core.py train -n MyVoice -sr 40k -f rmvpe -e contentvec -b 8 -te 300 -pg pretrained/G_40k.pth -pd pretrained/D_40k.pth

TTS with voice conversion:

python core.py tts -t "Hello, this is a test." -v "en-US-JennyNeural" -m logs/MyModel/G_1234.pth -id logs/MyModel/MyModel.index

Blend two models:

python core.py model_blender -m1 logs/ModelA/G_100.pth -m2 logs/ModelB/G_200.pth -o logs/Blended/G_blended.pth -r 0.6

Inspect model metadata:

python core.py model_information -m logs/MyModel/G_1234.pth