haoone-app: a local AI subtitle studio for DaVinci Resolve and Premiere Pro editors
新一代 AI 专业字幕软件，剪映字幕、elevenlabs 语音转文本的最佳本地版平替之一，也是加强版。中英转录识别准确率超过 97%，词语音频对齐率 98%，说话人分割与识别准确率 96%，带有最先进的 ASR 开源模型，100+字幕动画（支持导出透明背景的字幕动画）。说话人识别、专业字幕编辑器、命令行工具、Skill，达芬奇字幕插件（含字幕动画插件），PR 字幕插件，本地转录、远程转录、文稿匹配、智能拆行、AI校正、AI 智能热词、翻译、双语字幕、专业字幕编辑器、字幕合成、自定义大模型 API
At a glance
- What is it?
- haoone-app is a desktop subtitle application built around open ASR models rather than Whisper, with DaVinci and Premiere plugins, a CLI, and a paid one-time unlock for advanced features. Here is what the repository and README actually document, and where the gaps are.
- Who is it for?
- haoone-app fits editors who work in DaVinci Resolve or Premiere Pro and want local transcription with word-level alignment, bilingual subtitles, and a subtitle animation library without a subscription. It is a poor fit if you need a fully open source pipeline, since the app itself is distributed as a download and the licence is not stated in the repository.
- Can I use it commercially?
- Not without permission. GitHub finds no licence file in the repository, and without a licence all rights are reserved by default: you may read the code but not reuse it. Check the README, or ask the authors, before using it.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly MDX, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 29, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
What haoone-app is for, and who it is aimed at
haoone-app is a desktop subtitle application. The README describes it as a local alternative to Jianying (CapCut) subtitles and ElevenLabs speech-to-text, and it lists the target audience implicitly: people who cut video in DaVinci Resolve or Adobe Premiere Pro and need subtitles generated, aligned, edited, translated and burned in without leaving the editing workflow. The repository name and the guide links point to a GUI application, not a library. The README also points to separate repositories for a command line tool (haoone-cli) and an agent skill (haoone-skill), which suggests the desktop app is the hub and the CLI and skill are alternative entry points for automation.
The problem it claims to solve is specific. The README argues that many existing subtitle tools are wrappers around Whisper, and that Whisper has weaknesses in Chinese transcription and timestamp alignment. Whether or not that characterisation is fair, it explains the design: haoone-app integrates what the README calls the most advanced open source ASR models, chosen from the author's own evaluations, rather than bundling every available model. The README states that the model list is curated and updated over time, and links to evaluation videos. That curation is the product's main thesis: the value is in model selection, a Rust and C++ transcription engine, and a word-level alignment algorithm, not in the GUI alone.
How the transcription pipeline is put together
The README states that the transcription engine is built on Rust and C++, that GPU acceleration is enabled automatically, and that word-level audio alignment is developed in-house. It does not describe the internal data flow, so the architecture can only be inferred at a high level: audio goes in, a local or remote ASR model produces text, and an alignment stage maps words to time ranges. The README gives a word alignment rate of 98% for Chinese and English and 99% for Japanese, and a speaker diarisation accuracy of 96%.
Two transcription paths are documented. Local transcription runs on the machine and the README names qwen3-asr as the base for the local model, with multilingual support. Remote transcription is a hosted service, and the README claims accuracy above Jianying and the Doubao API at a lower price than Doubao. There is also support for a custom large model API, and the README says haoone is possibly the first subtitle application to support the mimo-v2.5-asr API, quoting a cost of 0.5 yuan for one hour of transcription.
The editing layer is where the project spends most of its feature list: AI line breaking, AI hotword replacement, AI correction, script matching with no length limit, bilingual subtitles, translation, search and batch replace, a fast edit view and a chapter view. The README notes that words replaced in the fast edit view automatically become hotwords. That feedback loop is the most concrete mechanism described: corrections feed back into recognition for later files.
Installing haoone-app and transcribing a first file
The repository does not contain install instructions for the application itself. It is not distributed through the repository; the README links to a download page, and the guide links to an installation and update page. The README does not document a package manager, a Docker image, or a source build, so the only documented path is downloading the desktop build for Windows or macOS from the project site.
What the repository does contain is the plugin and skill code. The top-level entries include haoone-pr-plugin/ and haoone-skill, and the README states that the PR subtitle plugin and the haoone-skill are open source, while the DaVinci subtitle plugin is still pending open source. For the PR plugin, the repository is the source of truth:
git clone https://github.com/minghe36/haoone-app.git
cd haoone-app/haoone-pr-pluginThe README does not list build or install commands for the plugin inside the repository, so after cloning, the guide page for the PR plugin is the place to look for the installation procedure. For the skill, the README describes it as usable in various agents to transcribe video and podcast subtitles and produce transcripts with high-precision timestamps. The README does not give a package name or an install command for the skill, so treat the haoone-skill repository as the entry point.
Once the desktop application is installed, the README describes the first workflow as opening a project, adding one or more media files, and running local transcription. The README states that a three-minute video completes in under 20 seconds on a Mac mini M4 with 16 GB of memory, and that a two-hour file takes about 10 minutes on Mac. These are the project's own figures, not independent measurements. Model download is handled inside the application, with one-click download and an alternative network drive download. After transcription, the fast edit view is where corrections happen, and replaced words become hotwords automatically.
Where haoone-app is the wrong tool
The biggest limitation is licensing and distribution. The repository does not state a licence, and the application is not built from this repository. If your organisation requires a known open source licence, an auditable build, or the ability to patch the transcription engine, this project does not offer that today. The engine is described as Rust and C++ but the engine source is not in the top-level entries listed here. The README says the DaVinci plugin is pending open source, which means the DaVinci integration is currently a binary you install, not code you can inspect.
The second limitation is platform. The README says Windows and macOS are supported. Android appears in search phrases around the project, but the repository does not document an Android build, and the download page is the only distribution channel mentioned. If you need a Linux workstation or a headless server pipeline, the desktop application is not the right shape; the CLI repository is the documented alternative, but this repository does not document its installation.
The third limitation is the remote transcription path. It depends on a hosted service, so it is not a fully local workflow, and the README does not document data handling for that service beyond a general statement that the software does not collect transcription data or media files. The README also does not document rollback for model updates or for application versions, so if a new model regresses on your material, the documented recovery path is not stated.
How haoone-app differs from Whisper-based subtitle tools
The README explicitly positions haoone-app against Whisper-based software. The stated difference is model choice and alignment. A typical Whisper wrapper uses one model family for all languages and accepts its segment-level timestamps. haoone-app instead uses a curated set of open ASR models, with a separate Japanese model that the README says reaches 94% recognition accuracy and 99% word-level alignment, and it applies its own word alignment stage on top. The README also claims better handling of Chinese and English mixed speech, proper nouns, songs and dialects than Whisper in 2026.
A second difference is the editing surface. Whisper-based tools often stop at producing a subtitle file. haoone-app bundles a subtitle editor with fast edit and chapter views, batch replace, hotword feedback, bilingual subtitles, translation, and over 100 subtitle animations including per-character and karaoke styles, with transparent background export. The README also documents DaVinci and Premiere plugins so transcription can run inside the NLE without opening the application.
A third difference is commercial. The README states that basic features are free forever and advanced features are a one-time purchase with no subscription. That is a different model from hosted transcription services billed per minute, and from fully free open source tools with no paid tier. The trade-off is that the paid tier is not itemised in the repository; the README links to a separate paid benefits page.
Maintenance, releases and upgrade cost
The repository is not archived, and the last push was on 2026-09-16. The most recent release listed is V13.4.0 on 2026-09-16, following V13.3.1 on 2026-09-14 and V13.3.0 on 2026-09-14. The README states that the project has iterated through 11 major versions and is under continuous iteration. The release cadence visible in the release list is rapid, with multiple releases in the same week, which is consistent with the README's claim of ongoing development.
Upgrade cost has two parts. The application itself is a download, and the README says basic features are permanently free while advanced features are a one-time buyout with no subscription. The repository does not list which features sit behind the paid tier or what the price is; the README links to a paid benefits page in the guide. Model updates are handled inside the application via one-click download, with a network drive alternative, so the operational cost of keeping models current is low, but the README does not document how to pin or roll back a model version.
Licence implications are unclear because the repository does not state a licence. The README says the PR plugin and the skill are open source and that the DaVinci plugin is pending open source, but it does not name the licence for any of them. If you plan to redistribute the plugins or embed the skill in a commercial product, the licence terms need to be confirmed from the individual repositories rather than assumed from this README.
Editorial conclusion
haoone-app fits editors who work in DaVinci Resolve or Premiere Pro and want local transcription with word-level alignment, bilingual subtitles, and a subtitle animation library without a subscription. It is a poor fit if you need a fully open source pipeline, since the app itself is distributed as a download and the licence is not stated in the repository. Before adopting it, verify the current download page, the paid feature list in the guide, and whether the DaVinci plugin has been open sourced, since the README still lists that as pending.
Frequently asked questions
Is haoone-app free to use?
The README states that basic features are free to use permanently, and that advanced features are a one-time purchase with no subscription fee. The repository does not itemise which features are in the paid tier; the README links to a separate paid benefits page in the guide.
Does haoone-app work on Windows and macOS?
The README says the software works on both Windows and macOS, and that it can be used even on machines with weaker performance. It does not document a Linux build or an Android build.
Can haoone-app transcribe without an internet connection?
Yes, local transcription is one of the two documented paths, based on qwen3-asr with in-house word-level alignment. The other path is remote transcription, which uses a hosted service and is described as faster and more accurate than Jianying and the Doubao API at a lower price.
Which editing applications does haoone-app integrate with?
The README documents a DaVinci Resolve subtitle plugin, including a subtitle animation plugin, and a Premiere Pro subtitle plugin. It states that the PR plugin is open source and that the DaVinci plugin is pending open source.
Does haoone-app collect my audio or transcription data?
The README states that the software does not collect users' transcription data or audio and video files. It does not describe the data handling of the remote transcription service beyond that statement.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/minghe36-haoone-app)
Community notes