LALAL.AI API v1

repository·master·Indexed 41 days ago

API for high-quality audio source separation, voice conversion, and background music removal. Features include a stem splitter for extracting up to 10 different instrument stems, a voice converter with accent and tonality control, and tools for removing echo, reverb, and background noise. Includes Python scripts (lalalai_splitter.py, lalalai_voice_converter.py, and lalalai_demuser.py) to facilitate integration with the asynchronous API workflow.

Tokens: 2.4K
Snippets: 5
Records: 11
Agent score: 42%

What's inside LALAL.AI

Use LALAL.AI Services via API
master
The LALAL.AI API provides access to several specialized audio processing services:
- Stem Splitter: Separates audio into up to 10 stems, including vocals, instrumentals, drums, bass, guitars, piano, synthesizer, wind, and string instruments.
- Voice Cloner: Enables the creation of custom voice clones.
- Voice Changer: Provides accent and tonality control with echo estimation and intelligent routing.
- Lead & Back Vocal Splitter: Isolates lead vocals from backing vocals.
- Echo & Reverb Remover: Removes unwanted echo and reverberation.
- Voice Cleaner: Removes background noise and extraneous sounds from recordings.
https://github.com/omnisalegmbh/lalalai/blob/master/readme.md
Understand the LALAL.AI API v1 Workflow
master
All LALAL.AI API v1 operations follow an asynchronous lifecycle:
1. Upload: Send your file to the /api/v1/upload/ endpoint to receive a source_id.
2. Process: Initiate a task (splitting or voice conversion) using the source_id.
3. Check: Periodically poll the /api/v1/check/ endpoint to monitor task status.
4. Download: Once the task is complete, use the provided URLs to download the results.
5. Delete (Optional): Use the /api/v1/delete/ endpoint to clean up files from LALAL.AI storage.
https://github.com/omnisalegmbh/lalalai/blob/master/api-v1/python/README.md
Install requirements for LALAL.AI Python examples
master
To use the provided Python scripts for interacting with the LALAL.AI API v1, you must install the requests library.
```
pip install requests
```
https://github.com/omnisalegmbh/lalalai/blob/master/api-v1/python/README.md
Use lalalai_demuser.py to clean voice from background music
master
The lalalai_demuser.py script is a specialized tool designed to isolate and clean a voice from background music.
```
python3 lalalai_demuser.py --license <user license> \
                             --input <input directory or file> \
                             [--output <output directory>]
```
https://github.com/omnisalegmbh/lalalai/blob/master/api-v0(deprecated)/tools/api/README.md

Use lalalai_demuser.py to remove background music

master

The lalalai_demuser.py script is a specialized tool to remove background music from voice recordings using the /api/v1/split/demuser/ endpoint.

python3 lalalai_demuser.py --license <user license> \
                           --input <input directory or file> \
                           [--output <output directory>] \
                           [--dereverb-enabled <true/false>] \
                           [--delete <true/false>]

Access the LALAL.AI API V1
master
LALAL.AI provides a stable API (V1) for integrating stem splitting and audio processing into your applications. It features a rich set of capabilities, an OpenAPI specification, and a Swagger-like interface for testing endpoints.
For Python developers, there are dedicated tools and scripts available in the api-v1/python/ directory of this repository to facilitate integration.
Key Resources:
- Human-readable documentation: https://www.lalal.ai/api/v1/docs/
- OpenAPI Specification: https://www.lalal.ai/api/v1/openapi.json
https://github.com/omnisalegmbh/lalalai/blob/master/readme.md
Use lalalai_voice_converter.py for voice conversion
master
The lalalai_voice_converter.py script converts a voice to match a specific voice pack.
Usage Modes:
- New Upload: Provide --input <file>.
- Existing Upload: Provide --uploaded_file_id <file id> to skip the upload step.
- List Voice Packs: Run with --list to see available voice packs and exit.
Key Parameters:
- --voice_pack_id: The ID of the target voice (default: ALEX_KAYE).
- --accent: Strength of accent matching (0.0 to 1.0). 0.0 keeps original; 1.0 matches target.
- --tonality-reference: Choose source_file to keep original pitch or voice_pack to match target pitch.
- --splitter: Choose a model (orion, perseus, phoenix, andromeda).
```
python3 lalalai_voice_converter.py --license <user license> \
                                   [--input <input directory or file>] \
                                   [--uploaded_file_id <file id>] \
                                   [--output <output directory>] \
                                   [--voice_pack_id <voice pack id>] \
                                   [--accent <0.0-1.0>] \
                                   [--tonality-reference <source_file/voice_pack>] \
                                   [--dereverb-enabled <true/false>] \
                                   [--splitter <model>] \
                                   [--delete <true/false>] \
                                   [--list]
```
https://github.com/omnisalegmbh/lalalai/blob/master/api-v1/python/README.md
Use lalalai_splitter.py for audio source separation
master
The lalalai_splitter.py script allows you to extract specific stems (like vocals, drums, or bass) from an audio source.
Key Stem Behaviors:
- voice: Uses the /split/voice_clean/ endpoint for voice cleaning. Supports --noise-cancelling (0, 1, or 2).
- music: Uses the /split/demuser/ endpoint to remove music from voice.
- All other stems (e.g., drum, bass, piano, guitar, synthesizer, strings, wind) use the /split/stem_separator/ endpoint.
Important Options:
- --extraction-level: For stem separator stems, choose between deep_extraction (more detail) or clear_cut (cleaner).
- --multivocal: For the vocals stem, use lead_back to split into lead and backing vocals.
- --splitter: Select a model (auto, orion, perseus, phoenix, andromeda, lynx, lyra). auto picks the latest available for the selected stem.
```
python3 lalalai_splitter.py --license <user license> \
                            --input <input directory or file> \
                            [--output <output directory>] \
                            [--stem <stem option>] \
                            [--noise-cancelling <level>] \
                            [--dereverb-enabled <true/false>] \
                            [--extraction-level <level>] \
                            [--multivocal <option>] \
                            [--splitter <model>] \
                            [--delete <true/false>]
```
https://github.com/omnisalegmbh/lalalai/blob/master/api-v1/python/README.md

Reference: lalalai_splitter.py CLI arguments

master

Arguments for the lalalai_splitter.py script:

Flag	Description	Default	Choices
`--license`	User license key (required)	N/A	N/A
`--input`	Input directory or file (required)	N/A	N/A
`--output`	Output directory	current script directory	N/A
`--stem`	Stem to extract	`vocals`	`vocals`, `drum`, `bass`, `piano`, `electric_guitar`, `acoustic_guitar`, `synthesizer`, `voice`, `strings`, `wind`
`--filter`	Post-processing filter intensity	`1`	`0` (mild), `1` (normal), `2` (aggressive)
`--splitter`	Neural network type	`auto`	`phoenix`, `orion`, `perseus`
`--enhanced-processing`	Enable enhanced processing	`false`	`true`, `false`
`--noise-cancelling`	Noise cancelling level (for `voice` stem only)	`1`	`0` (mild), `1` (normal), `2` (aggressive)

Reference: lalalai_voice_converter.py CLI arguments

master

Arguments for the lalalai_voice_converter.py script:

Flag	Description	Default	Choices/Range
`--license`	User license key (required)	N/A	N/A
`--input`	Input directory or file	N/A	N/A
`--uploaded_file_id`	Previously uploaded file ID	N/A	N/A
`--output`	Output directory	current script directory	N/A
`--voice_pack_id`	Voice pack ID	`ALEX_KAYE`	See API list
`--accent_enhance`	Accent processing strength	`1.0`	`0.0` to `1.0`
`--pitch_shifting`	Match target voice tonality	`true`	`true`, `false`
`--dereverb_enabled`	Remove echo/reverb	`false`	`true`, `false`
`--list`	List available voice packs and exit	N/A	N/A

LALAL.AI API v1 Authentication and Notes
master
When interacting with the LALAL.AI API v1:
- Authentication: All API requests must include your license key in the X-License-Key header.
- Asynchronicity: Processing tasks are not instantaneous. You must poll the /api/v1/check/ endpoint to determine when a task has finished.
- Output: Successful tasks return direct download URLs for the processed audio tracks.
https://github.com/omnisalegmbh/lalalai/blob/master/api-v1/python/README.md