January 22, 2026

Telegram + YouTube, your daily news podcast briefing

Lisa Granqvist Partner Workflow Automation Expert

Get a free AI assessment → ⬇️ Use template

Your morning “news routine” probably looks like this: ten tabs open, two YouTube videos half-watched, and a sinking feeling you still missed the one story that actually matters.

This Telegram news podcast automation hits busy marketers and founders hardest, but consultants and operators feel it too. You want a clean briefing you can listen to, not another feed to babysit.

This n8n workflow pulls from RSS and YouTube, filters the junk, summarizes with AI, then delivers both a readable recap and an MP3 to Telegram. Below, you’ll see how it works and what it replaces in your day.

How This Automation Works

The full n8n workflow, from trigger to final output:

n8n Workflow Template: Telegram + YouTube, your daily news podcast briefing

Click to explore

flowchart LR

    subgraph sg0["Schedule Flow"]
        direction LR
        n0@{ icon: "mdi:play-circle", form: "rounded", label: "Schedule Trigger", pos: "b", h: 48 }
        n1@{ icon: "mdi:cog", form: "rounded", label: "RSS Feed Read (TechCrunch)", pos: "b", h: 48 }
        n2@{ icon: "mdi:cog", form: "rounded", label: "RSS Feed Read (The Verge)", pos: "b", h: 48 }
        n3["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/merge.svg' width='40' height='40' /></div><br/>Merge - Tech (Append)"]
        n4["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Code - Clean & Dedup (newsCo.."]
        n5@{ icon: "mdi:cog", form: "rounded", label: "YouTube - Search (Latest 24h)", pos: "b", h: 48 }
        n6["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/merge.svg' width='40' height='40' /></div><br/>Merge - Add YouTube (Append)"]
        n7["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Code - YouTube Normalize + F.."]
        n8@{ icon: "mdi:swap-vertical", form: "rounded", label: "Set - Mark RSS", pos: "b", h: 48 }
        n9@{ icon: "mdi:swap-vertical", form: "rounded", label: "Set - Mark YouTube", pos: "b", h: 48 }
        n10@{ icon: "mdi:robot", form: "rounded", label: "Google Gemini Chat (AI Analy..", pos: "b", h: 48 }
        n11["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Code - Robust Parser (Gemini.."]
        n12["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/telegram.svg' width='40' height='40' /></div><br/>Telegram - Send Briefing Text"]
        n13["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/httprequest.dark.svg' width='40' height='40' /></div><br/>HTTP - OpenAI TTS (speech)"]
        n14["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/telegram.svg' width='40' height='40' /></div><br/>Telegram - Send Briefing Audio"]
        n15["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Code - Fix Audio Meta (filen.."]
        n16["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Code - Build TTS Payload (Op.."]
        n17@{ icon: "mdi:cog", form: "rounded", label: "RSS Feed Read (BBC World)", pos: "b", h: 48 }
        n18["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/merge.svg' width='40' height='40' /></div><br/>Merge - Add BBC World (Append)"]
        n8 --> n6
        n0 --> n1
        n0 --> n2
        n0 --> n17
        n0 --> n5
        n9 --> n6
        n3 --> n18
        n17 --> n18
        n2 --> n3
        n13 --> n15
        n1 --> n3
        n6 --> n4
        n5 --> n7
        n18 --> n8
        n10 --> n11
        n16 --> n13
        n7 --> n9
        n4 --> n10
        n11 --> n16
        n11 --> n12
        n15 --> n14
    end

    %% Styling
    classDef trigger fill:#e8f5e9,stroke:#388e3c,stroke-width:2px
    classDef ai fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
    classDef aiModel fill:#e8eaf6,stroke:#3f51b5,stroke-width:2px
    classDef decision fill:#fff8e1,stroke:#f9a825,stroke-width:2px
    classDef database fill:#fce4ec,stroke:#c2185b,stroke-width:2px
    classDef api fill:#fff3e0,stroke:#e65100,stroke-width:2px
    classDef code fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
    classDef disabled stroke-dasharray: 5 5,opacity: 0.5
    class n0 trigger
    class n10 ai
    class n13 api
    class n4,n7,n11,n15,n16 code
    classDef customIcon fill:none,stroke:none
    class n3,n4,n6,n7,n11,n12,n13,n14,n15,n16,n18 customIcon

The Problem: Morning News Becomes a Time Trap

Keeping up with the news sounds simple until you try to do it consistently. You check TechCrunch, then The Verge, then a world feed like BBC, then you bounce to YouTube because “someone will explain it faster.” Suddenly it’s 30 minutes, then 60, and you still don’t feel informed. You feel scattered. Worse, the content mix is noisy: duplicated headlines, thin rewrites, promo posts disguised as “analysis,” and videos that are interesting but not relevant to what you actually need today.

That friction compounds fast. Here’s where it usually breaks down in real life.

You spend your best morning energy sorting information instead of using it.
Duplicate stories show up everywhere, so you waste time rereading the same update in different words.
YouTube is valuable, but it’s a rabbit hole, and “quick research” turns into twenty minutes.
Even when you find good pieces, you don’t get a portable summary you can listen to while commuting or doing admin.

The Solution: A Daily Telegram Briefing (Text + MP3)

This workflow turns your news inputs into one small, reliable output. It runs on a schedule (7:00 AM by default), collects items from multiple RSS feeds like TechCrunch, The Verge, and BBC World, and also searches YouTube for recent videos around a topic you care about. Then it cleans everything up: duplicates removed, spammy or promotional items filtered, and YouTube results normalized so the AI sees a consistent structure. Google Gemini reads the combined set, selects the top stories, and rewrites them into a short, clear script. Finally, OpenAI Text-to-Speech converts that script into an MP3, and the workflow sends both the written briefing and the audio file straight to your Telegram chat.

The workflow starts with scheduled collection across RSS and YouTube. It then deduplicates and prepares the items for Gemini to summarize into a tight briefing. After that, OpenAI generates the audio, and Telegram delivers the text plus the MP3 so you can read or listen.

What You Get: Automation vs. Results

What This Workflow Automates

Results You’ll Get

It pulls fresh stories from multiple RSS feeds on a daily schedule.
It searches YouTube for recent videos tied to your chosen keyword.
It deduplicates, normalizes, and tags items so summaries stay clean.
It turns the final script into an MP3 and sends it to Telegram automatically.

Most people get about 30 minutes back every morning.
You stop “tab hopping” and start the day with one briefing.
The summary is easier to remember because it’s rewritten as a single narrative.
You can listen while commuting, exercising, or doing inbox cleanup.
Your briefings stay consistent day-to-day, which makes the habit stick.

Example: What This Looks Like

Say you normally check 3 sites and watch 2 YouTube videos each morning. If you spend about 10 minutes per site and 15 minutes per video, that’s roughly 60 minutes, and it’s easy to go longer. With this workflow, you spend maybe 2 minutes updating your topics when needed, then the 7:00 AM run delivers a Telegram summary plus an MP3 a little later. Your “news intake” becomes a 10-minute listen, not a one-hour scroll.

What You’ll Need

n8n instance (try n8n Cloud free)
Self-hosting option if you prefer (Hostinger works well)
Telegram to receive the text briefing and MP3.
YouTube Data API to search recent topic videos.
Google Gemini API key (get it from Google AI Studio / Google Cloud credentials).
OpenAI API key (get it from your OpenAI dashboard).
Telegram Bot Token (get it from BotFather in Telegram).

Skill level: Intermediate. You’ll paste API keys, set a Telegram chat ID, and test one full run end-to-end.

Don’t want to set this up yourself? Talk to an automation expert (free 15-minute consultation).

How It Works

A scheduled morning trigger runs automatically. The workflow kicks off at a set time (7:00 AM by default), so you’re not remembering to press buttons before coffee.

RSS feeds and YouTube are collected in parallel. n8n reads multiple RSS sources (TechCrunch, The Verge, BBC World in this template) and runs a YouTube search for recent videos tied to your topic keyword.

The items are cleaned up and shaped for summarization. Code steps remove duplicates and normalize YouTube results, then “tag” RSS and video items so Gemini can compare them properly instead of mixing formats.

AI creates a briefing, then audio is generated and delivered. Gemini produces a short script and summary text, OpenAI turns the script into an MP3, and Telegram sends both messages so you can read fast or hit play.

You can easily modify the RSS sources and YouTube keywords to match your niche, so the briefing stays useful instead of generic. See the full implementation guide below for customization options.

Step-by-Step Implementation Guide

Step 1: Configure the Schedule Trigger

Set the workflow to run on a daily schedule so it can pull fresh news and videos automatically.

Add and configure Scheduled Automation Trigger as the workflow trigger.
Set Rule to trigger at hour 7 using the interval schedule (as shown in the node configuration).
Confirm the workflow timezone aligns with your preferred run time (workflow settings show Asia/Ho_Chi_Minh).
Note the parallel execution: Scheduled Automation Trigger outputs to Fetch TechCrunch Feed, Retrieve Verge Feed, Collect BBC World Feed, and Search YouTube Recent in parallel.

Step 2: Connect RSS and YouTube Sources

Pull news from RSS feeds and recent YouTube videos, then merge the sources for downstream processing.

In Fetch TechCrunch Feed, set URL to https://techcrunch.com/feed/ and keep Ignore SSL enabled.
In Retrieve Verge Feed, set URL to https://www.theverge.com/rss/index.xml and keep Ignore SSL enabled.
In Collect BBC World Feed, set URL to http://feeds.bbci.co.uk/news/world/rss.xml and keep Ignore SSL enabled.
In Search YouTube Recent, set Resource to video, Limit to 15, and configure filters: q to AI, Tech News, Artificial Intelligence, regionCode to US, and publishedAfter to {{$now.setZone('America/New_York').minus({ hours: 24 }).toISO()}}.
Credential Required: Connect your youTubeOAuth2Api credentials in Search YouTube Recent.
Ensure Fetch TechCrunch Feed and Retrieve Verge Feed merge into Combine Tech Sources, then connect Combine Tech Sources and Collect BBC World Feed into Append World Feed.

Step 3: Normalize, Tag, and Deduplicate Content

Standardize RSS and YouTube formats, tag source types, and remove duplicates before AI analysis.

Configure Normalize YouTube Results to filter and format YouTube items (code uses a 24-hour cutoff and keyword filters).
In Tag YouTube Items, add a field named sourceType with value =youtube and enable Include Other Fields.
In Tag RSS Items, add a field named sourceType with value =rss and enable Include Other Fields.
Connect Tag YouTube Items and Tag RSS Items into Append YouTube Items, then connect Append YouTube Items to Cleanse And Deduplicate.
Review Cleanse And Deduplicate settings in the code: ONLY_LAST_HOURS = 36, MAX_ITEMS = 60, and MIX_RATIO of rss: 0.7 and youtube: 0.3.

Tip: The workflow uses multiple code nodes (5 total). If you customize filters or limits, keep the values consistent between Normalize YouTube Results and Cleanse And Deduplicate to avoid unexpected drops.

Step 4: Set Up AI Briefing Generation

Use Gemini to generate the briefing and parse it into Telegram-friendly formats.

In Gemini Briefing Analyzer, select Model models/gemini-2.5-flash and set Temperature to 0.3.
Set the system message in Gemini Briefing Analyzer to You are a professional News Editor. You must follow instructions exactly..
Ensure the prompt includes the news context and date expression: {{$json.newsContext}} and {{$now.setZone('America/New_York').toFormat('yyyy-LL-dd')}}.
Credential Required: Connect your googlePalmApi credentials in Gemini Briefing Analyzer.
Connect Gemini Briefing Analyzer to Parse Gemini Output to validate JSON and clean Telegram HTML.
Note the parallel execution: Parse Gemini Output outputs to both Build TTS Request and Post Telegram Briefing in parallel.

Step 5: Configure Telegram Text and Audio Outputs

Send the briefing as a Telegram post and generate an audio briefing using OpenAI TTS.

In Post Telegram Briefing, set Text to {{ $json.telegram_body }} and replace Chat ID with your Telegram ID; keep Parse Mode as HTML.
Credential Required: Connect your telegramApi credentials in Post Telegram Briefing.
In Build TTS Request, verify the TTS payload values: model gpt-4o-mini-tts, voice alloy, response_format mp3, and speed 1.05.
In OpenAI Speech Request, set URL to https://api.openai.com/v1/audio/speech and JSON Body to {{ $json.tts }}.
Credential Required: Connect your openAiApi credentials in OpenAI Speech Request.
In Adjust Audio Metadata, keep the filename logic to set morning_briefing_YYYY-MM-DD.mp3 and ensure the output has audio/mpeg mime type.
In Send Telegram Audio, set Operation to sendAudio, enable Binary Data, set Binary Property Name to audio, and set Caption to {{ $json.title }}.
Credential Required: Connect your telegramApi credentials in Send Telegram Audio.

⚠️ Common Pitfall: Telegram requires valid HTML. If Parse Gemini Output fails or Telegram rejects the message, check the cleaned HTML output for disallowed tags or malformed links.

Step 6: Test & Activate

Validate the full workflow from content retrieval through Telegram delivery before enabling automation.

Click Execute Workflow to run a manual test starting from Scheduled Automation Trigger.
Confirm merged feed items flow into Cleanse And Deduplicate and that Gemini Briefing Analyzer returns valid JSON.
Verify Post Telegram Briefing sends a formatted HTML message to your chat and Send Telegram Audio delivers an MP3 with the expected caption.
If everything succeeds, toggle the workflow to Active to enable scheduled delivery.

🔒

Unlock Full Step-by-Step Guide

Get the complete implementation guide + downloadable template

Common Gotchas

Telegram credentials and chat permissions are easy to get wrong. If messages don’t arrive, check the bot token and the chat ID used in the Telegram nodes first.
If you’re using Wait nodes or external rendering, processing times vary. Bump up the wait duration if downstream nodes fail on empty responses.
Default prompts in AI nodes are generic. Add your brand voice early or you’ll be editing outputs forever.

Frequently Asked Questions

How long does it take to set up this Telegram news podcast automation?

About 45 minutes if you already have the API keys.

Do I need coding skills to automate Telegram news podcast delivery?

No. You’ll mostly paste credentials and tweak a few fields in n8n. The code nodes are already built, so you’re configuring, not programming.

Is n8n free to use for this Telegram news podcast workflow?

Yes. n8n has a free self-hosted option and a free trial on n8n Cloud. Cloud plans start at $20/month for higher volume. You’ll also need to factor in OpenAI Text-to-Speech and Gemini usage, which depends on how long your daily script is.

Where can I host n8n to run this automation?

Two options: n8n Cloud (managed, easiest setup) or self-hosting on a VPS. For self-hosting, Hostinger VPS is affordable and handles n8n well. Self-hosting gives you unlimited executions but requires basic server management.

Can I customize this Telegram news podcast workflow for a different topic like real estate or crypto?

Yes, and it’s the most useful tweak you can make. Swap the RSS URLs in the RSS Read nodes to match your niche, then change the keywords in the YouTube search node so it pulls relevant videos. You can also adjust the “Gemini Briefing Analyzer” prompt to prioritize certain sources or formats (for example, “only include items with numbers or product announcements”). If you want a different sound, update the voice settings in the code step that builds the TTS payload.

Why is my Telegram connection failing in this workflow?

Usually it’s the chat ID.

How many items can this Telegram news podcast automation handle?

On n8n Cloud Starter, you’re usually fine for a daily run like this, since it’s one execution per morning plus the internal steps. If you self-host, there’s no execution limit, so the real constraint becomes your server size and how many RSS items and videos you pull each day. Practically, most teams keep it to a few feeds and a focused YouTube query so the AI summary stays short and listenable. If you want “everything,” expect higher AI costs and longer audio generation time.

Is this Telegram news podcast automation better than using Zapier or Make?

For a multi-step briefing like this, n8n is usually a better fit because you can run heavier logic (deduping, merging, parsing AI output) without paying extra for every branch. It’s also easier to self-host, which matters if you want more control over scheduled runs and costs. Zapier and Make can still do it, but you’ll often end up stitching together more scenarios and workarounds, especially around AI formatting and audio handling. If you only want “RSS in, Telegram message out,” those tools are fine. Talk to an automation expert if you want help choosing.

Set this up once and your mornings get quieter, honestly. The workflow handles the collecting, filtering, summarizing, and audio delivery so you can just listen and move on.

Need Help Setting This Up?

Our automation experts can build and customize this workflow for your specific needs. Free 15-minute consultation—no commitment required.

Lisa Granqvist

Workflow Automation Expert

Expert in workflow automation and no-code tools.

{
"id": null,
"meta": {
"instanceId": "867dee779e2cfbbacb7585a44f9f9418fb28b86c12f411ecf3cbd9ac454582af",
"templateCredsSetupCompleted": null,
"templateId": null
},
"name": "Automated News Digest Delivery",
"tags": [],
"nodes": [
{
"id": "flowpast-topbar-12141",
"name": "Flowpast Branding",
"type": "n8n-nodes-base.stickyNote",
"position": [
0,
20
],
"parameters": {
"color": 7,
"width": 1290,
"height": 80,
"content": "## Flowpast.com | Automation Workflow Library\n**\ud83d\udcd6 Full tutorial & setup guide:** flowpast.com"
},
"typeVersion": 1
},
{
"id": "a6b1fbdc-2d05-49b4-bf26-3c1dd6cf2d92",
"name": "Scheduled Automation Trigger",
"type": "n8n-nodes-base.scheduleTrigger",
"position": [
480,
120
],
"parameters": {
"rule": {
"interval": [
{
"triggerAtHour": 7
}
]
}
},
"typeVersion": 1.3
},
{
"id": "069ba213-9d37-4262-b79f-de447bdcf064",
"name": "Fetch TechCrunch Feed",
"type": "n8n-nodes-base.rssFeedRead",
"onError": "continueRegularOutput",
"position": [
0,
320
],
"parameters": {
"url": "https://techcrunch.com/feed/",
"options": {
"ignoreSSL": true
}
},
"retryOnFail": false,
"typeVersion": 1.2
},
{
"id": "4be9bbdd-8e45-4eab-bd19-a9d640730edc",
"name": "Retrieve Verge Feed",
"type": "n8n-nodes-base.rssFeedRead",
"onError": "continueRegularOutput",
"position": [
310,
230
],
"parameters": {
"url": "https://www.theverge.com/rss/index.xml",
"options": {
"ignoreSSL": true
}
},
"retryOnFail": false,
"typeVersion": 1.2
},
{
"id": "d699416f-d446-4c93-8cb6-cd3a82aec355",
"name": "Combine Tech Sources",
"type": "n8n-nodes-base.merge",
"position": [
190,
430
],
"parameters": [],
"typeVersion": 3.2
},
{
"id": "276f62c2-69a7-4f30-8fc0-c96d187720c5",
"name": "Cleanse And Deduplicate",
"type": "n8n-nodes-base.code",
"position": [
460,
690
],
"parameters": {
"jsCode": "// ===== CONFIG =====\nconst ONLY_LAST_HOURS = 36;          // Filter news from the last X hours (set null/0 if not filtering)\nconst MAX_ITEMS = 60;    // total number of items included in context\nconst MAX_TOTAL_CHARS = 28000;   // Block token overshoot (optional)\nconst MAX_TITLE_CHARS = 160;\nconst MAX_DESC_CHARS = 260;\n\n// Mix ratio by source type (RSS vs YouTube)\n// Example 70/30 => MAX_ITEMS=60 => RSS ~42, YouTube ~18 (if there's a shortage, it will be automatically compensated)\nconst MIX_RATIO = {\n  rss: 0.7,\n  youtube: 0.3,\n};\n\n// Remove posts that are in the form of deals/prices/where to buy... (less clutter and less length)\nconst EXCLUDE_IF_TEXT_MATCH = [\n  /\\bwhere to buy\\b/i,\n  /\\bdeal\\b/i,\n  /\\bsale\\b/i,\n  /\\bpromo code\\b/i,\n  /\\$\\d/i,\n  /\\bblack friday\\b/i,\n  /\\bcyber monday\\b/i,\n];\n\n// The Verge's signature phrase is often inserted.\nconst STRIP_PHRASES = [\n  /read the full story at the verge\\.?/gi,\n  /read our review\\.?/gi,\n];\n\n// ==================\n\nfunction decodeEntities(str) {\n  str = String(str ?? '');\n\n  const map = {\n    ' ': ' ',\n    '&': '&',\n    '"': '\"',\n    ''': \"'\",\n    ''': \"'\",\n    '<': '<',\n    '>': '>',\n    '’': '\u2019',\n    '‘': '\u2018',\n    '“': '\u201c',\n    '”': '\u201d',\n    '—': '\u2014',\n    '–': '\u2013',\n    '…': '\u2026',\n  };\n\n  str = str.replace(\n    / |&|"|'|'|<|>|’|‘|“|”|—|–|…/g,\n    (m) => map[m] ?? m\n  );\n\n  // numeric entities: ’  /  ’\n  str = str.replace(/&#(\\d+);/g, (_, n) => {\n    const code = parseInt(n, 10);\n    return Number.isFinite(code) ? String.fromCodePoint(code) : _;\n  });\n  str = str.replace(/&#x([0-9a-fA-F]+);/g, (_, h) => {\n    const code = parseInt(h, 16);\n    return Number.isFinite(code) ? String.fromCodePoint(code) : _;\n  });\n\n  return str;\n}\n\nfunction stripHtml(s) {\n  return decodeEntities(String(s ?? ''))\n    .replace(/<[^>]*>/g, ' ')\n    .replace(/\\s+/g, ' ')\n    .trim();\n}\n\nfunction clip(s, n) {\n  s = String(s ?? '').trim();\n  if (!n || s.length <= n) return s;\n  return s.slice(0, n - 1).trimEnd() + '\u2026';\n}\n\nfunction normTitle(s) {\n  return stripHtml(s).toLowerCase().replace(/\\s+/g, ' ').trim();\n}\n\nfunction inferPublication(link, feedTitle) {\n  const l = String(link ?? '');\n  const ft = String(feedTitle ?? '');\n  if (/techcrunch\\.com/i.test(l)) return 'TechCrunch';\n  if (/theverge\\.com/i.test(l)) return 'The Verge';\n  if (/vnexpress\\.net/i.test(l)) return 'VNExpress';\n  if (/youtube\\.com|youtu\\.be/i.test(l)) return 'YouTube';\n  if (/trends\\.google\\.com/i.test(l)) return 'Google Trends';\n  if (ft) return stripHtml(ft);\n  return 'RSS';\n}\n\nfunction inferSourceType(j) {\n  // Prioritize fields set from upstream.\n  if (j?.sourceType) return String(j.sourceType);\n\n  const link = String(j?.link || j?.url || '');\n  if (/youtube\\.com|youtu\\.be/i.test(link)) return 'youtube';\n  if (/trends\\.google\\.com/i.test(link)) return 'trends';\n\n  return 'rss';\n}\n\nfunction parseDateMaybe(v) {\n  if (!v) return null;\n  const d = new Date(v);\n  return Number.isNaN(d.getTime()) ? null : d;\n}\n\nfunction matchesAny(reList, text) {\n  const t = String(text ?? '');\n  return reList.some((re) => re.test(t));\n}\n\nfunction getTs(a) {\n  const d = a.publishedAt ? new Date(a.publishedAt) : null;\n  const t = d && !Number.isNaN(d.getTime()) ? d.getTime() : 0;\n  return t;\n}\n\n// time cutoff\nconst now = new Date();\nconst cutoff = ONLY_LAST_HOURS ? new Date(now.getTime() - ONLY_LAST_HOURS * 3600 * 1000) : null;\n\n// main\nconst seen = new Set();\nconst candidates = [];\n\nlet droppedDup = 0;\nlet droppedByTime = 0;\nlet droppedByExclude = 0;\n\nfor (const it of $input.all()) {\n  const j = it.json || {};\n\n  const title = stripHtml(j.title || j.Title || '');\n  if (!title) continue;\n\n  const link = stripHtml(j.link || j.url || '');\n  const descRaw = stripHtml(j.description || j.content || j.snippet || '');\n  const author = stripHtml(j.creator || j.author || j.dcCreator || '');\n\n  // PublishedAt can be pubDate/isoDate/publishedAt...\n  const publishedAtRaw =\n    j.isoDate || j.pubDate || j.publishedAt || j.publishTime || j.date || '';\n  const publishedAt = parseDateMaybe(publishedAtRaw);\n\n  if (cutoff && publishedAt && publishedAt < cutoff) {\n    droppedByTime++;\n    continue;\n  }\n\n  // Filter out \"deals/junk\" based on aggregated text.\n  const haystack = `${title}\\n${descRaw}\\n${link}`;\n  if (matchesAny(EXCLUDE_IF_TEXT_MATCH, haystack)) {\n    droppedByExclude++;\n    continue;\n  }\n\n  // key dedup\n  const key = link ? `L:${link}` : `T:${normTitle(title)}`;\n  if (seen.has(key)) {\n    droppedDup++;\n    continue;\n  }\n  seen.add(key);\n\n  // clean desc\n  let desc = descRaw;\n  for (const re of STRIP_PHRASES) desc = desc.replace(re, '').trim();\n  desc = clip(desc, MAX_DESC_CHARS);\n\n  const publication = inferPublication(link, j.feedTitle);\n  const sourceType = inferSourceType(j);\n  const titleClipped = clip(title, MAX_TITLE_CHARS);\n\n  candidates.push({\n    title: titleClipped,\n    description: desc,\n    link,\n    publication,\n    author,\n    publishedAt: publishedAt\n      ? publishedAt.toISOString()\n      : (publishedAtRaw ? String(publishedAtRaw) : ''),\n    sourceType,\n  });\n}\n\n// ===== MIX by ratio (70/30) =====\n\n// Group by sourceType\nconst rssList = [];\nconst ytList = [];\nconst otherList = [];\n\nfor (const a of candidates) {\n  if (a.sourceType === 'youtube') ytList.push(a);\n  else if (a.sourceType === 'rss') rssList.push(a);\n  else otherList.push(a);\n}\n\n// Sort each group from newest to first (this is up to you; if you want to keep the input order, remove the sort).\nrssList.sort((a, b) => getTs(b) - getTs(a));\nytList.sort((a, b) => getTs(b) - getTs(a));\notherList.sort((a, b) => getTs(b) - getTs(a));\n\n// Quotas\nlet rssQuota = Math.round(MAX_ITEMS * (MIX_RATIO.rss ?? 0.7));\nrssQuota = Math.max(0, Math.min(MAX_ITEMS, rssQuota));\nlet ytQuota = MAX_ITEMS - rssQuota;\n\n// Pick quota\nlet pickedRss = rssList.slice(0, rssQuota);\nlet pickedYt = ytList.slice(0, ytQuota);\n\n// Fill n\u1ebfu 1 b\u00ean thi\u1ebfu\nif (pickedYt.length < ytQuota) {\n  const need = ytQuota - pickedYt.length;\n  pickedRss = pickedRss.concat(rssList.slice(rssQuota, rssQuota + need));\n}\nif (pickedRss.length < rssQuota) {\n  const need = rssQuota - pickedRss.length;\n  pickedYt = pickedYt.concat(ytList.slice(ytQuota, ytQuota + need));\n}\n\n// Alternate mixing according to the 70/30 pattern (7 RSS, 3 YT).\nconst w = 10;\nconst wR = Math.max(0, Math.min(10, Math.round((MIX_RATIO.rss ?? 0.7) * w)));\nconst wY = w - wR;\nconst pattern = [\n  ...Array(wR).fill('rss'),\n  ...Array(wY).fill('youtube'),\n];\n\nconst queueR = [...pickedRss];\nconst queueY = [...pickedYt];\n\nconst articles = [];\nlet pi = 0;\n\nwhile (articles.length < MAX_ITEMS && (queueR.length || queueY.length)) {\n  const want = pattern[pi % pattern.length];\n  pi++;\n\n  if (want === 'youtube' && queueY.length) articles.push(queueY.shift());\n  else if (want === 'rss' && queueR.length) articles.push(queueR.shift());\n  else if (queueR.length) articles.push(queueR.shift());\n  else if (queueY.length) articles.push(queueY.shift());\n}\n\n// (Optional) If still missing, fill with otherList\nif (articles.length < MAX_ITEMS && otherList.length) {\n  articles.push(...otherList.slice(0, MAX_ITEMS - articles.length));\n}\n\n// ===== build newsContext (for Gemini) =====\nlet newsContext = '';\nfor (const a of articles) {\n  // format: [Publication | Author | sourceType] Title: Desc (Link)\n  // (You can remove sourceType from the head if you don't want it to be visible)\n  const headParts = [a.publication];\n  if (a.author) headParts.push(a.author);\n  if (a.sourceType) headParts.push(a.sourceType);\n\n  const head = `[${headParts.join(' | ')}]`;\n\n  const line =\n    `${head} ${a.title}` +\n    `${a.description ? `: ${a.description}` : ''}` +\n    `${a.link ? ` (${a.link})` : ''}\\n`;\n\n  if ((newsContext.length + line.length) > MAX_TOTAL_CHARS) break;\n  newsContext += line;\n}\n\nreturn [{\n  json: {\n    newsContext: newsContext.trim(),\n    articles,\n    stats: {\n      inputCount: $input.all().length,\n      candidatesCount: candidates.length,\n      rssCount: rssList.length,\n      youtubeCount: ytList.length,\n      outputCount: articles.length,\n      droppedDup,\n      droppedByTime,\n      droppedByExclude,\n      cutoffISO: cutoff ? cutoff.toISOString() : null,\n      ratio: MIX_RATIO,\n      quota: {\n        rssQuota,\n        ytQuota,\n        pickedRss: pickedRss.length,\n        pickedYt: pickedYt.length,\n      }\n    }\n  }\n}];"
},
"typeVersion": 2
},
{
"id": "fab7cfbf-1b3f-4579-9111-239af2f8da4e",
"name": "Search YouTube Recent",
"type": "n8n-nodes-base.youTube",
"position": [
835,
340
],
"parameters": {
"limit": 15,
"filters": {
"q": "AI, Tech News, Artificial Intelligence",
"regionCode": "US",
"publishedAfter": "={{ $now.setZone('America/New_York').minus({ hours: 24 }).toISO() }}"
},
"options": {
"order": "date"
},
"resource": "video"
},
"credentials": {
"youTubeOAuth2Api": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1
},
{
"id": "2531e648-ab2a-4dc7-b130-7f30872843ac",
"name": "Append YouTube Items",
"type": "n8n-nodes-base.merge",
"position": [
605,
655
],
"parameters": [],
"typeVersion": 3.2
},
{
"id": "cef6ff91-7995-4e39-8a65-1ec9b38a3b1a",
"name": "Normalize YouTube Results",
"type": "n8n-nodes-base.code",
"position": [
780,
500
],
"parameters": {
"jsCode": "// ====== CONFIG ======\nconst ONLY_LAST_HOURS = 24;\n\n// Keyword \"Positive\" (Must match at least one)\nconst INCLUDE_REGEX = [\n  /\\bai\\b/i,\n  /\\bartificial intelligence\\b/i,\n  /\\btech\\b/i,\n  /\\bllm\\b/i,\n  /\\bchatgpt\\b/i,\n  /\\bgemini\\b/i,\n  /\\bopenai\\b/i,\n  /\\bautomation\\b/i,\n  /\\brobotics\\b/i,\n  /\\bmachine learning\\b/i,\n];\n\n// Keyword \"Negative\" (If matched, exclude item)\nconst EXCLUDE_REGEX = [\n  /\\bcrypto\\b/i,\n  /\\bbitcoin\\b/i,\n  /\\bcoin\\b/i,\n  /\\binvestment\\b/i,\n  /\\bprice prediction\\b/i,\n  /\\bxrp\\b/i,\n  /\\bgiveaway\\b/i,\n  /\\blive stream\\b/i, // Often low quality or just looping music\n];\n\n// ======================================\n\nconst now = new Date();\nconst cutoff = new Date(now.getTime() - ONLY_LAST_HOURS * 60 * 60 * 1000);\n\nfunction pickThumb(snippet) {\n  return (\n    snippet?.thumbnails?.high?.url ||\n    snippet?.thumbnails?.medium?.url ||\n    snippet?.thumbnails?.default?.url ||\n    ''\n  );\n}\n\nfunction matchAny(regexList, text) {\n  return regexList.some((re) => re.test(text));\n}\n\nconst out = [];\n\nfor (const item of $input.all()) {\n  const j = item.json;\n\n  const videoId = j?.id?.videoId || '';\n  const snippet = j?.snippet || {};\n\n  const title = snippet?.title || '';\n  const description = snippet?.description || '';\n  const channelTitle = snippet?.channelTitle || '';\n  const publishedAt = snippet?.publishedAt || snippet?.publishTime || '';\n\n  if (!videoId || !title) continue;\n\n  // Time filter (if publishedAt parse is possible)\n  if (publishedAt) {\n    const d = new Date(publishedAt);\n    if (!Number.isNaN(d.getTime()) && d < cutoff) continue;\n  }\n\n  // Content filter\n  const haystack = `${title}\\n${description}\\n${channelTitle}`;\n  if (!matchAny(INCLUDE_REGEX, haystack)) continue;\n  if (matchAny(EXCLUDE_REGEX, haystack)) continue;\n\n  out.push({\n    json: {\n      title,\n      link: `https://www.youtube.com/watch?v=${videoId}`,\n      description,\n      source: `YouTube - ${channelTitle || 'Unknown'}`,\n      publishedAt,\n      thumbnail: pickThumb(snippet),\n      channelTitle,\n      videoId,\n    }\n  });\n}\n\nreturn out;"
},
"typeVersion": 2
},
{
"id": "1c33a22a-aa9f-4cd2-bc01-6c033e15ea05",
"name": "Tag RSS Items",
"type": "n8n-nodes-base.set",
"position": [
465,
545
],
"parameters": {
"options": [],
"assignments": {
"assignments": [
{
"id": "2ef4eb06-d6fb-4ccb-a5c7-24e10fd4734e",
"name": "sourceType",
"type": "string",
"value": "=rss"
}
]
},
"includeOtherFields": true
},
"typeVersion": 3.4
},
{
"id": "27f323b2-af06-4321-a498-fbe4e46fe432",
"name": "Tag YouTube Items",
"type": "n8n-nodes-base.set",
"position": [
850,
545
],
"parameters": {
"options": [],
"assignments": {
"assignments": [
{
"id": "7fddd203-3144-4d14-af88-6608b0b5778c",
"name": "sourceType",
"type": "string",
"value": "=youtube"
}
]
},
"includeOtherFields": true
},
"typeVersion": 3.4
},
{
"id": "a84e7155-1591-4000-94a2-2982a116f3c3",
"name": "Gemini Briefing Analyzer",
"type": "@n8n/n8n-nodes-langchain.googleGemini",
"position": [
190,
845
],
"parameters": {
"modelId": {
"__rl": true,
"mode": "list",
"value": "models/gemini-2.5-flash",
"cachedResultName": "models/gemini-2.5-flash"
},
"options": {
"temperature": 0.3,
"systemMessage": "=You are a professional News Editor. You must follow instructions exactly."
},
"messages": {
"values": [
{
"content": "=Return ONLY valid JSON. No markdown fences. No extra text.\nRead this context:\n{{$json.newsContext}}\n\nTasks:\n1) Select top 10 most important stories (mix of Tech & Global News).\n2) Write a morning briefing in English.\n\nOutput strict JSON (no markdown fences, no extra text):\n{\n  \"title\": \"{{$now.setZone('America/New_York').toFormat('yyyy-LL-dd')}} Morning Briefing\",\n  \"telegram_body\": \"Use ONLY HTML: <b>, <i>, <a href='url'>. NO Markdown symbols (*, _, [). Keep under 4000 chars.\",\n  \"audio_script\": \"Natural, energetic English text for speech. No special chars like < > [ ] * _. Avoid URLs.\"\n}\n\nRules:\n- telegram_body must be valid Telegram HTML.\n- Use bullet points with '\u2022' not '-' if you want lists.\n- Always include links via <a href='...'>...</a>."
}
]
},
"simplify": false,
"jsonOutput": true
},
"credentials": {
"googlePalmApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1
},
{
"id": "b5e21b3a-2f50-4437-aeb9-297d685319e1",
"name": "Parse Gemini Output",
"type": "n8n-nodes-base.code",
"position": [
450,
950
],
"parameters": {
"jsCode": "const item = $input.first().json;\n\n// 1) Get raw text from Gemini\nconst raw = (item?.candidates?.[0]?.content?.parts || [])\n  .map(p => p?.text ?? '')\n  .join('')\n  .trim();\n\nif (!raw) throw new Error('Gemini output error: candidates[0].content.parts[].text not found');\n\n// 2) Cut JSON objects (to avoid any extra text).\nconst first = raw.indexOf('{');\nconst last = raw.lastIndexOf('}');\nif (first === -1 || last === -1 || last <= first) {\n  throw new Error('Not found JSON object in output Gemini');\n}\nconst jsonStr = raw.slice(first, last + 1);\n\n// 3) parse\nlet parsed;\ntry {\n  parsed = JSON.parse(jsonStr);\n} catch (e) {\n  throw new Error('JSON.parse failed. JSON string was:\\n' + jsonStr);\n}\n\n// 4) CLEAN HTML FOR TELEGRAM\nlet telegram_body = String(parsed.telegram_body ?? '').trim();\n\n// Fix 1: Replace the tags <br>, <br/>, <br /> with the newline character \\n\ntelegram_body = telegram_body.replace(/<br\\s*\\/?>/gi, '\\n');\n\n// Fix 2: Replace the <p> tag with a blank line and </p> with a line break (as a precaution).\ntelegram_body = telegram_body.replace(/<p>/gi, '').replace(/<\\/p>/gi, '\\n');\n\n// Fix 3: Normalize href (change ' to \" if applicable)\ntelegram_body = telegram_body.replace(/href='([^']+)'/g, 'href=\"$1\"');\n\nconst audio_script = String(parsed.audio_script ?? '').trim();\nconst title = String(parsed.title ?? '').trim();\n\nif (!telegram_body) throw new Error('Missing telegram_body');\nif (!audio_script) throw new Error('Missing audio_script');\n\nreturn [{\n  json: { title, telegram_body, audio_script }\n}];"
},
"typeVersion": 2
},
{
"id": "acecf9dc-e3f6-4e55-b69c-9f45b6b6dbcd",
"name": "Post Telegram Briefing",
"type": "n8n-nodes-base.telegram",
"position": [
150,
1060
],
"webhookId": "25fe2177-437e-4692-8589-0b49b777597f",
"parameters": {
"text": "={{ $json.telegram_body }}",
"chatId": "[YOUR_ID]",
"additionalFields": {
"parse_mode": "HTML"
}
},
"credentials": {
"telegramApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1.2
},
{
"id": "a8477a99-b8a2-412b-8413-2ab509743f06",
"name": "OpenAI Speech Request",
"type": "n8n-nodes-base.httpRequest",
"position": [
830,
850
],
"parameters": {
"url": "https://api.openai.com/v1/audio/speech",
"method": "POST",
"options": {
"response": {
"response": {
"responseFormat": "file",
"outputPropertyName": "audio"
}
}
},
"jsonBody": "={{ $json.tts }}",
"sendBody": true,
"specifyBody": "json",
"authentication": "predefinedCredentialType",
"nodeCredentialType": "openAiApi"
},
"credentials": {
"openAiApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 4.3
},
{
"id": "98377e3d-ec07-4b77-ad1a-fc7dc3106ffe",
"name": "Send Telegram Audio",
"type": "n8n-nodes-base.telegram",
"position": [
1190,
885
],
"webhookId": "a5717da9-c0e1-4129-84f1-ad085c443542",
"parameters": {
"chatId": "[YOUR_ID]",
"operation": "sendAudio",
"binaryData": true,
"additionalFields": {
"caption": "={{ $json.title }}"
},
"binaryPropertyName": "audio"
},
"credentials": {
"telegramApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1.2
},
{
"id": "e6b67eae-8d2f-45e4-aab3-c7f3e1d29dfc",
"name": "Adjust Audio Metadata",
"type": "n8n-nodes-base.code",
"position": [
1060,
860
],
"parameters": {
"jsCode": "const item = $input.first();\nif (!item.binary?.audio) {\n  throw new Error('Not found binary.audio from node HTTP TTS');\n}\n\nconst desiredNameFromJson = item.json.fileName || item.json.filename || item.json.audioFileName;\nlet desiredName = desiredNameFromJson || 'morning_briefing.mp3';\n\n// Make sure to add the .mp3 extension.\nif (!desiredName.toLowerCase().endsWith('.mp3')) desiredName += '.mp3';\n\n// Reset metadata for the binary.\nitem.binary.audio.fileName = desiredName;\n\n// mimeType: OpenAI usually returns audio/mpeg (correct). Telegram is also OK.\n// But if you want \"standard mp3\", then audio/mpeg is still the best option.\nitem.binary.audio.mimeType = 'audio/mpeg';\n\n// (Optional) some flows/nodes can read file extensions.\nitem.binary.audio.fileExtension = 'mp3';\n\nreturn [item];"
},
"typeVersion": 2
},
{
"id": "62b3256d-d14b-43c8-a9d5-a0daa724d841",
"name": "Build TTS Request",
"type": "n8n-nodes-base.code",
"position": [
705,
800
],
"parameters": {
"jsCode": "const { title, telegram_body, audio_script } = $input.first().json;\n\n// Guard + API-based restrictions (safe)\nlet input = String(audio_script || '').trim();\nif (!input) throw new Error('audio_script r\u1ed7ng');\n\n// OpenAI TTS input is limited to 4096 characters -> temporarily truncated to avoid failure.\nif (input.length > 4096) {\n  input = input.slice(0, 4050).trimEnd() + '...';\n}\n\nconst now = new Date();\nconst yyyy = now.getFullYear();\nconst mm = String(now.getMonth() + 1).padStart(2, '0');\nconst dd = String(now.getDate()).padStart(2, '0');\n\nreturn [{\n  json: {\n    title,\n    telegram_body,\n    fileName: `morning_briefing_${yyyy}-${mm}-${dd}.mp3`,\n    tts: {\n      model: 'gpt-4o-mini-tts',\n      voice: 'alloy',\n      input,\n      instructions: 'Clear, professional English broadcast voice.',\n      response_format: 'mp3',\n      speed: 1.05,\n    },\n  }\n}];"
},
"typeVersion": 2
},
{
"id": "6fb4d594-8bbb-42b8-aff2-24e74dec3f96",
"name": "Collect BBC World Feed",
"type": "n8n-nodes-base.rssFeedRead",
"onError": "continueRegularOutput",
"position": [
510,
330
],
"parameters": {
"url": "http://feeds.bbci.co.uk/news/world/rss.xml",
"options": {
"ignoreSSL": true
}
},
"retryOnFail": false,
"typeVersion": 1.2
},
{
"id": "d4caa68b-01fd-4c6c-8c0c-841c49ab99db",
"name": "Append World Feed",
"type": "n8n-nodes-base.merge",
"position": [
395,
400
],
"parameters": [],
"typeVersion": 3.2
}
],
"active": false,
"pinData": [],
"settings": {
"timezone": "Asia/Ho_Chi_Minh",
"callerPolicy": "workflowsFromSameOwner",
"timeSavedMode": "fixed",
"availableInMCP": false,
"executionOrder": "v1"
},
"versionId": null,
"connections": {
"Tag RSS Items": {
"main": [
[
{
"node": "Append YouTube Items",
"type": "main",
"index": 0
}
]
]
},
"Scheduled Automation Trigger": {
"main": [
[
{
"node": "Fetch TechCrunch Feed",
"type": "main",
"index": 0
},
{
"node": "Retrieve Verge Feed",
"type": "main",
"index": 0
},
{
"node": "Collect BBC World Feed",
"type": "main",
"index": 0
},
{
"node": "Search YouTube Recent",
"type": "main",
"index": 0
}
]
]
},
"Tag YouTube Items": {
"main": [
[
{
"node": "Append YouTube Items",
"type": "main",
"index": 1
}
]
]
},
"Combine Tech Sources": {
"main": [
[
{
"node": "Append World Feed",
"type": "main",
"index": 0
}
]
]
},
"Collect BBC World Feed": {
"main": [
[
{
"node": "Append World Feed",
"type": "main",
"index": 1
}
]
]
},
"Retrieve Verge Feed": {
"main": [
[
{
"node": "Combine Tech Sources",
"type": "main",
"index": 1
}
]
]
},
"OpenAI Speech Request": {
"main": [
[
{
"node": "Adjust Audio Metadata",
"type": "main",
"index": 0
}
]
]
},
"Fetch TechCrunch Feed": {
"main": [
[
{
"node": "Combine Tech Sources",
"type": "main",
"index": 0
}
]
]
},
"Append YouTube Items": {
"main": [
[
{
"node": "Cleanse And Deduplicate",
"type": "main",
"index": 0
}
]
]
},
"Post Telegram Briefing": {
"main": [
[]
]
},
"Search YouTube Recent": {
"main": [
[
{
"node": "Normalize YouTube Results",
"type": "main",
"index": 0
}
]
]
},
"Append World Feed": {
"main": [
[
{
"node": "Tag RSS Items",
"type": "main",
"index": 0
}
]
]
},
"Gemini Briefing Analyzer": {
"main": [
[
{
"node": "Parse Gemini Output",
"type": "main",
"index": 0
}
]
]
},
"Build TTS Request": {
"main": [
[
{
"node": "OpenAI Speech Request",
"type": "main",
"index": 0
}
]
]
},
"Normalize YouTube Results": {
"main": [
[
{
"node": "Tag YouTube Items",
"type": "main",
"index": 0
}
]
]
},
"Cleanse And Deduplicate": {
"main": [
[
{
"node": "Gemini Briefing Analyzer",
"type": "main",
"index": 0
}
]
]
},
"Parse Gemini Output": {
"main": [
[
{
"node": "Build TTS Request",
"type": "main",
"index": 0
},
{
"node": "Post Telegram Briefing",
"type": "main",
"index": 0
}
]
]
},
"Adjust Audio Metadata": {
"main": [
[
{
"node": "Send Telegram Audio",
"type": "main",
"index": 0
}
]
]
}
}
}