Bonsai-27B-GGUF
lmstudio-community · Kit
A GGUF conversion of the 27B Bonsai model for LM Studio, with a 3.80 GB Q1_0 language file and a 0.93 GB BF16 vision projector. Tested for text, vision and par…
An n8n webhook for YouTube links. It looks the video up, has Gemini watch it and write a structured summary, returns the summary to the caller and sends the title and link to Telegram.
Webhook that summarizes a YouTube video with Gemini is an n8n workflow in the Research category, listed by Mindket. It is free. This is version 1, last updated on . Download it on this page or copy its JSON, then import it into an n8n canvas. Nothing runs until you activate it in n8n.
Webhook
It starts when something posts a JSON body to the webhook:
A structured markdown summary of the video, returned to the caller, plus a Telegram message with the title and the link
What you need before you start.
Video reading and summary
Watches the YouTube video from its address and writes the structured summary. The file asks for gemini-3.8-flash.
NeedsA Gemini API key, kept in a Header Auth credential
Video details
Looks the video up for its title, description and id.
NeedsA YouTube OAuth2 credential, with the YouTube Data API on
Message
Sends the video’s title and link to a chat.
NeedsA Telegram bot credential and a chat id
Checked by Mindket on 13 Nov 2026
This n8n workflow turns a YouTube link into a structured summary. You post the link to a webhook. The workflow pulls the video’s id out of it and looks the video up on YouTube for its title and description. It then sends the video’s address to Google’s Gemini API, which watches the video and writes a summary under markdown headers. The summary goes back to the caller, and Telegram gets the title and the link. The file holds nine nodes, one of them a sticky note that Mindket added.
Researchers, students and content teams who want the gist of a long video before they sit through it. Skip it if you need the speaker’s exact words. The summary is the model’s reading of the video, and no transcript is kept.
| Industry | Who uses it | How they use it | Example |
|---|---|---|---|
| Research | Analysts | Skim a recorded talk before they decide to cite it | An analyst posts a conference talk and reads the headers before the full video. |
| Education | Students | Get a structured outline of a lecture | A student posts a recorded lecture and gets topics under headers with the key terms in bold. |
| Media | Content teams | Log what each source video covers | A newsroom assistant posts an interview and gets a summary and the title in a Telegram chat. |
| Software | Developers | Add video summaries to their own tools | A developer calls the webhook from a bookmark app that saves YouTube links. |
You need some n8n and three accounts: import a file, make three credentials and post a link to a webhook.
Any n8n that can reach YouTube, Google’s Gemini API and Telegram will run it.
Download saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. The original copy needed a community node for transcripts. It is no longer on npm, so Mindket swapped it for a Gemini call.
| Need | Detail |
|---|---|
| n8n | On n8n Cloud or on a server of your own |
| Gemini | An API key from Google AI Studio, kept in a Header Auth credential |
| Google Cloud | The YouTube Data API on, and an OAuth2 credential for the video lookup |
| Telegram | A bot, and the id of the chat it writes to |
workflow.json.x-goog-api-key and your Gemini key as the value. Select it in Summarize and Analyze Video.Get YouTube Video.Telegram and replace <YOUR_TELEGRAM_CHAT_ID> with your chat’s id. Delete the node if you do not want the message.Post a JSON body with one field, youtubeUrl. The call returns when the summary is ready, which can take a while for a long video. Gemini reads public videos only.
| To change | Where |
|---|---|
| The shape of the summary | The prompt inside the body of Summarize and Analyze Video |
| The model | The model line in the same body |
| What the caller gets back | The fields of Response Object |
| The Telegram message | The text of Telegram |
You post this to the webhook.
{"youtubeUrl": "https://www.youtube.com/watch?v=<VIDEO_ID>"}The call answers with a JSON object. summary holds the markdown summary, and title, description, id and youtubeUrl describe the video. Telegram receives the title and the link.
If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. This one runs in a straight line from the webhook to the answer, with Telegram on a side branch at the end. A webhook is an address that starts the run when something posts to it. Test with a short video first.
For experts. One HTTP Request node calls Gemini’s Interactions API, and the answer’s steps hold the text.
| Node | What it does | What to know |
|---|---|---|
YouTube Video ID | Code node that matches the id in a link | It reads watch, youtu.be and embed links, and a Shorts link gives no id |
Get YouTube Video | YouTube node, get one video | It needs a YouTube OAuth2 credential |
Summarize and Analyze Video | POST to /v1beta/interactions, gemini-3.8-flash, store false | The key goes in the x-goog-api-key header |
Response Object | Set node that picks the model’s text from steps | It fails when the answer has no steps |
The webhook has no authentication, and Gemini’s YouTube link feature is a preview.
Webhook before you publishstore is false, so Google keeps no copy of the exchange for later turns< or &topics is always an empty listAn n8n webhook for YouTube links. You post a link, the workflow looks the video up on YouTube, and Gemini watches it and
writes a structured summary. The summary goes back to the caller, and the title and link go to Telegram.
workflow.json: the workflow as n8n exports it, nine nodes. One is a sticky note that Mindket added. It holds noREADME.md: these notes.n8n-nodes-youtube-transcription, which turned the videoSummarize and Analyze Video. It sends the video's address to Google's Gemini API, which reads the video itself, soGet YouTube Video and the transcript node were not joined to anything after them, so a runGet YouTube Video now joins the new node.Telegram had no chat set. It holds the placeholder <YOUR_TELEGRAM_CHAT_ID>.Get YouTube Video needs.Telegram if you do not want the message.workflow.json.x-goog-api-key and whose value is your Gemini key. Select it inSummarize and Analyze Video.Get YouTube Video.Telegram and replace <YOUR_TELEGRAM_CHAT_ID> with the chat's id.Webhook and use its test address. Post a link to it:curl -X POST "<YOUR_N8N_HOST>/webhook-test/ytube" -H "Content-Type: application/json" -d '{"youtubeUrl": "https://www.youtube.com/watch?v=<VIDEO_ID>"}'
<YOUR_N8N_HOST>/webhook/ytube.Webhook takes a POST at the path ytube and keeps the caller waiting until Respond to Webhook.Get YouTube URL copies body.youtubeUrl into a field named youtubeUrl.YouTube Video ID pulls the 11-character id out of the link with a regular expression. It reads watch, youtu.beGet YouTube Video looks the video up, which gives the title, the description and the id.Summarize and Analyze Video posts to Gemini's interactions endpoint with the model gemini-3.8-flash. The bodystore is false, so Google keeps no copy of theResponse Object takes the model's text from the steps of the answer and sets summary, an empty topics,title, description, id and youtubeUrl.Respond to Webhook returns those fields to the caller as JSON, and Telegram sends the title and the link.Summarize and Analyze Video.model line in the same body.Response Object.Telegram.store is false.Webhook before you publish.< or &.topics is always an empty list. Nothing fills it.steps and Response Object fails.NOTICE.md 640 B README.md 5 KB workflow.json 8 KB
lmstudio-community · Kit
A GGUF conversion of the 27B Bonsai model for LM Studio, with a 3.80 GB Q1_0 language file and a 0.93 GB BF16 vision projector. Tested for text, vision and par…
LibertAIDAI · Kit
Runs DeepSeek-V4.1-Flash-1017 as a 2.9 bits-per-weight EXL3 quantization, a 196 GiB copy for people serving large models on their own hardware. Perplexity stay…
facebook · Kit
Generates short music clips from a text description or an audio prompt, in the small checkpoint of facebook's MusicGen. It suits developers and music hobbyists…
AbdomenAtlas · Kit
Detects and outlines tumors in 9 organs on CT scans, trained with the R-Super method that learns from radiology reports. A research model for medical imaging g…
dBenf · Kit
Answers typed questions about satellite image patches from all 12 Sentinel-2 bands, returning a probability for each option without writing text. A Qwen3.5-0.8…
zimmerk4 · Kit
Plays the full game Getting Over It from the bottom to the end with a small trained policy network that learned inside a JAX physics simulator. For reinforceme…