MockingBird
babysor/MockingBirdReviewed by AI
Checked by RepoPicks Radar on
What is MockingBird?
MockingBird copies the sound of a voice from a recording and then reads text out loud in that voice. It is built for Mandarin Chinese: the project says the original English model does not fit Chinese letters, so you either train your own voice model on large Chinese speech datasets or download one of the ready models that community members shared on file-sharing links. You run it from the terminal with Python, PyTorch and ffmpeg, and it then opens a page at http://localhost:8080 or a desktop window. An Nvidia graphics card is expected.
MockingBird is free and open source (MIT) and can replace ElevenLabs and Fish Audio.
How easy is it to install MockingBird?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- macOS E
- Linux D
Windows and Linux are the least hard and set the grade, but it is still a terminal job: install Python, PyTorch with the right CUDA version and ffmpeg, run "pip install -r requirements.txt", find a voice model, then start "python web.py" or "python demo_toolbox.py". On a Mac with Apple chip the README only offers a workaround: a Rosetta terminal, building pyworld and ctc-segmentation from source code and running Python through an x86 wrapper script, which is developer work.
Coming soon: install in 1 click with the RepoPicks app.
MockingBird on GitHub
- Stars
- 36,876on Nov 10, 2026
- Downloads
- —all releases
- Growth
- —no gain in the last reading
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Python, FFmpeg, NVIDIA graphics card
- To build from source
- Python, Docker, NVIDIA graphics card, FFmpeg
- Disk space
- 150 MB–450 MB · estimate
- Accounts or keys
- None
- Cost
- Free
- License
- MIT (Permissive)
- Can replace
- ElevenLabs, Fish Audio
- Last change on GitHub
- Mar 3, 2026
MockingBird risk signals (10)
- Attentionstale releases
Last release on 2021-11-07 (1823 days ago)
automatic check
- AttentionDownload links to file-sharing sites
README links to https://drive.google.com/file/d/110A11KZoVe7vy6kXlLb6zVPLb_J91I_t/view?usp=sharing
automatic check
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): numpy: GHSA-6p56-wp2h-9hxr, GHSA-fpfv-jqm9-f5jm; numpy: GHSA-6p56-wp2h-9hxr, GHSA-fpfv-jqm9-f5jm
automatic check
- AttentionLow OpenSSF Scorecard
OpenSSF Scorecard 2.5/10
automatic check
- Attentioncopies a real persons voice
The project describes itself as "Clone a voice in 5 seconds to generate arbitrary speech in real-time", and gen_voice.py takes a text file plus a .wav of the voice. Nothing asks whether the person agreed or marks the result as generated. Putting words in someone else's voice is restricted or punishable in several countries.
AI review
- Attentionvoice models come from file sharing links
Section 2.3 of README.md lists the ready voice models only as links posted by community members, with passwords, on pan.baidu.com, aliyundrive.com and a personal OneDrive (for example "https://pan.baidu.com/s/1iONvRxmkI-t1nHqxKytY3g 4j5d"). Three of them say "only works under version 0.0.1". These are PyTorch model files from strangers, loaded by the program.
AI review
- Attentionauthor stopped and points to a paid service
The first line of README.md: "While I no longer actively update this repo, you can find me continuously pushing this tech forward... I'm also building an optimized and cloud hosted version: https://noiz.ai/". The last release is from 2021 and the open issue count is 481.
AI review
- Infoonly works with old versions
README.md: "The recommended environment here is Repo Tag 0.0.1, Pytorch1.9.0 with Torchvision0.10.0 and cudatoolkit10.2 ... because requirements.txt was exported a few months ago, so it doesn't work with newer versions". requirements.txt still pins numpy==1.19.3 on Windows and numpy==1.20.3 elsewhere, the versions the checks flag for known vulnerabilities.
AI review
- Infochinese only and cli demo broken
The features list Mandarin only, the project is a fork of Real-Time-Voice-Cloning "which only support English", and section 2 warns "the demo_cli is not working at this moment, so additional synthesizer models are required". Without a Chinese model there is nothing ready to try.
AI review
- Infotraining needs a big machine
Training uses datasets such as aidatatang_200zh and aishell3 (tens of gigabytes of audio), docker-compose.yml reserves an Nvidia device, and FAQ item 6 tells Windows users to "change the virtual memory to 100G (102400)".
AI review
Evidence is shown in its original language.
What Radar could not check about MockingBird
- How good the cloned voice really sounds, and whether any of the shared models still load with today's PyTorch
- What is inside the model files on the Baidu and OneDrive links, since the project does not publish checksums
- Whether the install works at all with current Python and PyTorch instead of the 2021 versions
- Real disk space and training time, which depend on the Chinese datasets you download
- Whether it can be made to work for languages other than Mandarin