How to use VidBee: download, transcribe and summarize media

What VidBee does and who it is for

VidBee turns video and audio into an organized, searchable library. You can download media from supported sites such as YouTube, TikTok, Twitter, Instagram, Facebook, Twitch and Bilibili, or import local files. Its transcripts include speaker labels and timestamped playback.

• Saving videos or podcasts from supported sites into one library
• Searching lectures, interviews or podcast episodes by spoken text or speaker name
• Following channels through RSS and downloading new items automatically

Get VidBee and start the app

The project's website offers a download page for the desktop app at vidbee.org/download. The README does not list any prerequisites. It also includes a git clone command if you want the source code.

If you run the web and API version from source, the README gives these commands: pnpm run start:web starts apps/api and apps/web together, and docker compose up -d --build runs the Docker setup.

• Desktop app: download from the project's download page
• Source code: clone the repository from GitHub
• Web and API version: run from source or with Docker

Download video or audio

Paste one link or add a batch, then manage everything from the same desktop queue. The queue shows active, completed and failed downloads, with live progress, speed, file size and the format being saved.

For video, you can choose Auto (MP4/MKV), MP4, MKV, WebM or Original, or save audio separately. Use one-click defaults for a fast download, or choose formats for a specific item.

• Pause, resume, retry or remove tasks without rebuilding the queue
• Playlists and channels can also be saved from supported sites

Transcribe a download or a local file

Create a transcript from a finished download, or import a local audio or video file. Speech recognition runs on your computer, so the media does not need to be uploaded to a transcription service.

You can use source captions when they are available, or run local speech recognition for a new transcript. Choose from Whisper, SenseVoice, Parakeet and Qwen3-ASR model families. VidBee can recommend a model for your computer and language, and you can download or switch models from Settings.

• Detect speakers automatically, or set a fixed speaker count and re-label the conversation without changing the recognized words
• Search spoken text or speaker names, then click a timestamp to continue playback from that moment
• Copy the transcript, export plain text or Markdown, or create a new video with selectable or burned-in captions

Run AI prompts on a transcript

Run AI prompts against the transcript without leaving VidBee. Built-in prompts can create bullet summaries, clean up grammar, generate FAQs, extract statistics, build mind maps, paraphrase text or translate the transcript.

You choose the provider. VidBee can connect to OpenAI, Anthropic, Google, DeepSeek, Groq, Azure, Hugging Face, OpenRouter, xAI, Ollama or LM Studio, and you can add a custom provider with a Base URL, model ID and API key. When you run an AI prompt, the prompt and transcript content go to the provider you selected. Choose Ollama or LM Studio to keep that step on your computer too.

• Test a custom provider's connection before using it
• Create your own prompts with a title, icon and instructions
• API keys are stored on your computer

Automate downloads with RSS subscriptions

Subscribe to RSS feeds and VidBee checks for new items and adds them to your download queue. Each subscription can use its own rules, so regular releases do not need the same manual setup each time.

You can turn automatic downloads on or keep new items for manual review. Filters, tags, a custom directory and a filename template can be set per subscription.

• Filter feed items by keywords
• Download only the latest item if you do not want the backlog
• See whether each feed item is queued, downloading, completed, failed or waiting