HaizFlow
A free Windows app for downloading, translating, subtitling and dubbing videos.
Installation · User guide · Help · Tiếng Việt
About HaizFlow
HaizFlow helps you download videos, translate speech, create subtitles and add translated voice tracks. Process one video or a batch, cover original subtitles and assign different voices to different speakers. You can publish processed videos to social media through Zernio.
Automatic mode runs the steps you select. The Manual editor lets you review and correct captions or voice tracks before exporting.
Main features
| Feature | What you can do |
|---|---|
| Batch video downloads | Select multiple videos from supported public channels or profiles and queue them for download. |
| Video translation | Recognize speech with Whisper; translate locally with HY-MT2 or use Gemini. Review translations before exporting. |
| Subtitle creation | Create captions from video speech; edit text, timing, Bangers styling and karaoke highlighting. |
| Video dubbing | Generate speech for translated text using built-in voices or a reference sample you have permission to use. |
| Automatic processing | Choose languages, models and voice settings, then run your selected recognition, translation, dubbing and caption-coverage steps. |
| Batch processing | Apply common settings to multiple videos and follow each video's progress and results. |
| Cover original subtitles | Detect caption regions with OCR and cover them using blur or background patching. |
| Multiple-speaker detection | Group speakers and assign a separate voice to each group instead of using one voice for the entire conversation. |
| Social publishing | Select an account, prepare the post and publish through Zernio. You can use processed videos directly from projects. |
Manual tools run independently: dubbing is optional when you only need translated captions.
Preview the video, edit captions and choose suitable voices before exporting or publishing.
App screenshots
Home · Automatic processing · Manual editor · Downloads · Social publishing · Settings
The screenshots show the Vietnamese interface. Options and version labels may differ in newer releases. Click an image to view it at full size.
Home
Find recent projects, filter by project type and create a new project.
Automatic processing
Choose recognition, translation and voice models, then configure subtitle coverage, audio, background music and watermark settings.
Manual editor
Review source and result previews, run individual processing steps, split the Result timeline into clips, import SRT captions and export original or translated SRT for other editors.
Downloads
Inspect a video, channel or audio link, review the channel's video list and select the items to download. This example shows a Douyin channel.
Social publishing
Add processed videos from existing projects to the publishing queue and prepare posts for your connected account through Zernio.
Settings
Choose the interface language and CPU/GPU mode. The other tabs manage API keys and resource packs.
Local translation without per-call API charges
Whisper and HY-MT2 run on your computer after their resource packs are installed. They do not require API keys or per-call translation fees. Gemini is an additional option for online translation.
Computers without an NVIDIA GPU can use supported CPU options. Compatible NVIDIA GPUs can use GPU models. Choose a processing mode in Settings and install the matching resources for the tools you want to use.
Gemini, Zernio and other external services have their own terms, limits and charges. OmniVoice has the usage restriction described below.
Install and get started
Download the Windows installer. Run this EXE only; do not download or extract Core files. Resource packs are installed inside HaizFlow. Version 1.0.0 can update directly in the app. Release and installation details.
- Install and open HaizFlow. Vietnamese is the default; English is available in Settings.
- Choose CPU or NVIDIA GPU in Settings → General, then install the resource packs you need.
- Create an Automatic, Manual or Batch project and add videos.
- Select language, models and voice if needed. Process, review and Export the video.
Supported systems are Windows 10 version 1809 or later and Windows 11 x64, with at least 16 GB system RAM. Installer users do not need Python. Resource packs, projects and exports need additional storage; Setup and the Resource packs page show their corresponding requirements.
GPU mode supports compatible NVIDIA 4 GB-class cards (at least 3.5 GiB dedicated VRAM reported by CUDA). On 4 GB, use Whisper Small GPU, HY-MT2 CPU Q4 or Gemini, and OmniVoice GPU; these heavy stages run sequentially, with one voice sentence at a time. Whisper Turbo and full HY-MT2 GPU need more VRAM. Available VRAM and system memory still matter at runtime. Windows Smart App Control or an Application Control policy may block this unsigned build; see the installation guide.
See Installation for CPU/GPU packs, API keys and storage. See the User guide for projects, captions, dubbing and publishing.
License and commercial use
HaizFlow is free to use under HaizFlow Source-Available 1.0. This is not an OSI open-source license. Redistribution, repackaging and commercial software conditions are specified in LICENSE.
The current OmniVoice model is licensed for noncommercial use only. Do not assume it can be used for advertising, monetized videos or client work without appropriate permission. HaizFlow's license does not replace the model license.
You need permission to use imported video, music and voice samples. See NOTICE and third-party notices. The licensing review and inactive draft terms are retained separately.
HaizFlow uses Qt/PySide under LGPL and FFmpeg under GPL/LGPL, depending on the component. Bundled library sources and library replacement instructions are provided separately. Independent library-license rights remain available.
Support and contributions
See Help for resources, API keys and data management. You can also send feedback or request support. Never include keys or private data.
Source installation, testing and packaging are covered in the development guide. See the technical documentation to contribute code.
Created by Mạch Hồng Hải (Mach Hong Hai). Star the repository to follow the project.