New V4 Released!!!

New features:

• Add support for images! Now the program can transcode images to AVIF just like videos before.
• Fix various display issues & providing more details.
• Rebuild the queue part for new image function.

Features

• With WebUI, convenient for checking progress and operation.
• Run ffmpeg locally, without performance loss.
• Support full video normalization parameters, including pixel format and color space.
• Task queue, with automatic hang-up interrupt.
• With task schedule, easily plan your transcoding time.
• Bulk import & Global settings.
• GBM task prediction, more precise and lightly.
• AAutomatic Speech Recognition (ASR) by whisper.cpp and translation by llm (support openai API, mlx and llama).utomatic Speech Recognition & Translation.

Upgrade to V4

In the all-new version 4, the database structure and program data path has been changed.

Thus, if you want to upgrade to v4, the recommanded steps are totally uninstall the V3 and do a completely reinstall.

If you truly want to save the history (which used for ETA Model), you can do an manully update.

1. Find the config.db.
2. Delete all the tables except history.
3. Rename the history table to video_history.
4. Move the config.db to data/.

Installations

0. Install Python

Important

We do not recommend using 3.14.7 on MacOS as the task pause&resume function will face an uncoverable crash. See here.

This doesn't effect Windows. Not tested on Linux.

1. Install FFmpeg

Follow the instruction from ffmpeg.org.

2. Install Api

• Download the latest release.
• Unzip the file:

tar -xzf backend.tar.gz

• Install pip packages:

uv sync

• Run Api:

uv run python api/main.py

3. Install Docker

Docker Hub

docker run -d -p 8889:80 starstreammm/stvav1ui:latest

Github

docker run -d -p 8889:80 ghcr.io/starstreammm/svtav1ui:latest

4. Install libomp

The XGBoost need libomp to function. Some distributions/operating systems may include this component by default.

MacOS

brew install libomp

Linux/Windows

Search for it.

5. (Optional) Install Whisper.cpp

The ASR function is base on it.

To install, follow the instructions at Whisper.cpp.

6. (Optional) Install LLM framework

You can choose one from openai API, mlx and llama.

For OpenAI API,

uv pip install openai

For MLX (only recommendated for Apple Silicon),

uv pip install mlx-lm

For llama.cpp,

uv pip install llama-cpp-python

Links

FFmpeg: https://ffmpeg.org

SVT-AV1 Encoder: https://gitlab.com/AOMediaCodec/SVT-AV1/

MUI: https://mui.com/material-ui/