Healthcare patient education video creation from clinical guidelines

Creating patient education videos is one of those tasks that sits awkwardly between clinical importance and production tedium. Your medical team has evidence-based guidelines, clinical protocols, and patient education materials that could genuinely help people understand their conditions and treatments. Yet turning those documents into engaging, on-brand videos requires coordinating multiple specialists: scriptwriters, voice actors, video editors, and animators. Most healthcare organisations either skip this step entirely or pay thousands per video....... For more on this, see Healthcare clinic patient education video creation. For more on this, see Healthcare patient education video pipeline. For more on this, see AI Tools for Freelance Video Editors: Creating Content 3x....

The real bottleneck is the handoff problem. Someone writes the script. Someone else records the voiceover. A third person builds the video template. Then it sits in email chains while you wait for approvals. By the time a video is ready, your guidelines have been updated twice.

What if you could feed your clinical guidelines directly into an automated system and have a finished, branded patient education video appear on the other end? No email chains, no manual coordination, no three-week turnaround. This Alchemy workflow does exactly that: it takes a PDF or text-based guideline, generates a patient-friendly script, creates a professional voiceover, and builds a video featuring a realistic human presenter, all in one unbroken chain.

The Automated Workflow

Why this tool combination works

AI-Boost handles the clinical-to-patient translation brilliantly. It understands medical language and can rewrite dense protocols into something comprehensible without losing accuracy. ElevenLabs provides the voice: you choose from dozens of realistic voices and can select different ones for different patient demographics. Hour One is the orchestration glue on the video side; it creates videos with AI-generated presenters, perfect for healthcare where you need a consistent, trustworthy human face without the logistics of hiring and scheduling actual presenters.

The three tools have different strengths, and the orchestration layer ties them together. We'll use n8n here because it offers the most flexible webhook support and best handles the longer processing times these AI services require.

Step 1: Trigger and guideline extraction

Start with a webhook that listens for new guidelines. This could be triggered manually through an n8n interface, or you could point it at a document storage system like Google Drive or SharePoint. For this example, we'll assume the guideline arrives as a text upload.

Copy
POST /webhook/patient-education-trigger
Content-Type: application/json

{
  "guideline_text": "Type 2 Diabetes Management Protocol v4.2...",
  "condition_name": "Type 2 Diabetes",
  "target_audience": "newly diagnosed adults",
  "tone": "reassuring and practical",
  "video_language": "en-GB"
}

In n8n, create a webhook node that accepts this payload. Store the guideline text in a variable for the next step.

Step 2: Script generation with AI-Boost

AI-Boost's API accepts clinical text and returns patient-friendly versions. You'll need your API key from the AI-Boost dashboard.

Copy
POST https://api.ai-boost.com/v1/transform
Authorization: Bearer YOUR_AI_BOOST_KEY
Content-Type: application/json

{
  "input_text": "Type 2 Diabetes is a metabolic disorder characterised by insulin resistance...",
  "source_tone": "clinical",
  "target_tone": "patient education",
  "target_reading_level": "secondary school",
  "output_format": "video script",
  "max_length": 250,
  "include_cta": true
}

The response includes a structured script with scene markers:

Copy
{
  "script": {
    "intro": "Hello, I'm Dr Sarah. Today we're going to talk about type 2 diabetes...",
    "main_sections": [
      {
        "title": "What is Type 2 Diabetes?",
        "duration_seconds": 45,
        "text": "Your body produces a hormone called insulin..."
      },
      {
        "title": "How You Can Help",
        "duration_seconds": 60,
        "text": "The good news is you have real control here..."
      }
    ],
    "closing": "Talk to your GP about your personal plan...",
    "total_duration_seconds": 180,
    "reading_time_seconds": 195
  }
}

In n8n, add an HTTP Request node pointing to the AI-Boost endpoint. Map your webhook payload's guideline_text to the input_text field. Store the response script for step 3.

Step 3: Voiceover generation with ElevenLabs

Take the cleaned script and send it to ElevenLabs. You'll want to split longer scripts into sections to manage API limits (ElevenLabs has per-request character limits).

Copy
POST https://api.elevenlabs.io/v1/text-to-speech
xi-api-key: YOUR_ELEVENLABS_KEY
Content-Type: application/json

{
  "text": "Hello, I'm Dr Sarah. Today we're going to talk about type 2 diabetes. Your body produces a hormone called insulin, which helps glucose enter your cells...",
  "model_id": "eleven_monolingual_v1",
  "voice_settings": {
    "stability": 0.5,
    "similarity_boost": 0.75
  },
  "voice_id": "EXAVITQu4vr4xnSDxMaL"
}

The voice_id above is a female British voice, suitable for healthcare content. ElevenLabs returns audio as binary data. In n8n, use a File Write node to save the audio stream. For multi-section scripts, you'll need to loop through each section, call the API for each, and store all audio files.

Example of looping in n8n: use a Split node to iterate through your script sections, call ElevenLabs for each, and collect the results.

Step 4: Video composition with Hour One

Hour One takes a script and voiceover and produces a video with an AI-generated presenter. Their API expects you to specify avatar details, background, and optionally supply pre-recorded audio.

Copy
POST https://api.hourone.com/api/videos
Authorization: Bearer YOUR_HOUR_ONE_KEY
Content-Type: application/json

{
  "script_text": "Hello, I'm Dr Sarah. Today we're going to talk about type 2 diabetes...",
  "avatar": {
    "type": "avatar_id",
    "avatar_id": "emily_professional"
  },
  "voice": {
    "type": "external_audio",
    "audio_url": "https://your-storage.s3.amazonaws.com/voice_section_1.mp3"
  },
  "background": {
    "type": "image_url",
    "image_url": "https://your-brand-assets.com/healthcare_background.jpg"
  },
  "video_quality": "1080p",
  "output_format": "mp4"
}

Hour One's processing is asynchronous. The API returns a job ID immediately. You then poll the status endpoint to check progress.

Copy
GET https://api.hourone.com/api/videos/{job_id}
Authorization: Bearer YOUR_HOUR_ONE_KEY

Response while processing:

Copy
{
  "job_id": "vid_abc123xyz",
  "status": "processing",
  "progress": 65,
  "estimated_completion_seconds": 45
}

Response when complete:

Copy
{
  "job_id": "vid_abc123xyz",
  "status": "completed",
  "output_url": "https://videos.hourone.com/vid_abc123xyz.mp4",
  "duration_seconds": 195,
  "created_at": "2024-01-15T14:32:00Z"
}

In n8n, add an HTTP Request node for the initial POST, then use a Wait node set to poll every 10 seconds. Add a second HTTP Request node to check status, with a condition: if status is "completed", continue to the next step; if "processing", wait another 10 seconds.

Step 5: Storage and notification

Once the video is ready, you probably want it stored somewhere accessible to your team, and someone needs to know it's ready. Add nodes to:

1. Download the video from Hour One's URL and save to your storage (S3, Google Drive, or SharePoint).

2. Send a notification: email to your clinical team, Slack message to a channel, or webhook to your internal system.

Copy
POST https://hooks.slack.com/services/YOUR/SLACK/WEBHOOK
Content-Type: application/json

{
  "text": "Patient education video ready",
  "blocks": [
    {
      "type": "section",
      "text": {
        "type": "mrkdwn",
        "text": "Your Type 2 Diabetes video is ready for review.\nCondition: Type 2 Diabetes\nDuration: 3 minutes 15 seconds\n<https://your-storage.com/video.mp4|View video>"
      }
    }
  ]
}

Full n8n workflow structure

1. Webhook Trigger node (listens for guideline submission)
2. HTTP Request node (calls AI-Boost)
3. Split node (breaks script into sections for parallel processing)
4. Loop: HTTP Request node (calls ElevenLabs for each section)
5. Loop: File Write node (saves audio files)
6. HTTP Request node (initiates Hour One video generation)
7. Wait node (polls for completion)
8. HTTP Request node (checks completion status)
9. File Write node (downloads final video)
10. Slack/Email node (notifies team)

This handles the entire pipeline with zero manual intervention once you submit the guideline.

The Manual Alternative

If you prefer more control over each step, you can run these tools individually and keep the outputs in your existing workflow. Submit the guideline to AI-Boost through their web interface and download the script. Copy-paste the script into ElevenLabs' dashboard, select your voice, download the audio. Upload both to Hour One manually through their web app, select your avatar and background, and download the final video.

This approach takes perhaps 20 minutes per video and gives you a chance to review and adjust between steps. It's sensible if you're producing fewer than three or four videos per week, or if your guidelines frequently need custom tweaking that the AI isn't quite capturing correctly. For higher volume or standardised guidelines, the automation is worth the initial setup effort.

Pro Tips

Handling API rate limits: ElevenLabs and Hour One both impose rate limits on free and starter plans. If you're generating multiple videos, stagger requests using n8n's delay nodes. For example, set a 5-second delay between ElevenLabs calls. This prevents hitting rate limit errors and keeps your costs predictable.

Voice consistency: ElevenLabs voice IDs are stable, so if you choose "Emily Professional" for your first video, stick with it across all videos in a series. Your audience will recognise the voice and associate it with your trusted medical content. Test different voice IDs on a short script first; the difference between voices is more noticeable in healthcare contexts than marketing content.

Script length and video pacing: Hour One works best with scripts between 150 and 300 words (roughly 2 to 4 minutes of video). Longer scripts feel slow; shorter ones feel rushed. Test your first video at around 180 words and adjust based on what feels natural. AI-Boost's output_format parameter set to "video script" already optimises for this, but always check the returned duration estimate.

Audio file format: ElevenLabs returns audio in MP3 by default. Hour One accepts MP3, WAV, and OGG. MP3 is the safest choice because it's widely supported and has reasonable file sizes. Store audio files in an accessible location (S3, Wasabi, or your own CDN) rather than locally on your machine; Hour One's API needs a publicly accessible URL.

Monitoring and error recovery: In n8n, add error handling nodes after the AI-Boost and Hour One calls. If AI-Boost fails (for example, if your guideline text is malformed), catch the error and send a notification to your team with details. If Hour One fails partway through video generation, your workflow stops and you can retry. Set up automatic retries for transient failures (network timeouts), but alert the team for persistent errors (invalid voice ID, malformed script).

Cost Breakdown

This cost scales linearly with video volume. If you're producing 20 videos per month, ElevenLabs usage increases significantly. At that volume, consider ElevenLabs' Pro or Enterprise plan. Hour One's Pro plan is capped at 30 videos monthly; beyond that, you'd move to their Business tier (pricing on request).

For a typical NHS trust or healthcare provider producing 5–8 videos per month, the all-in cost is roughly £400–500 monthly, or £50–60 per video. This is cheaper than hiring freelance video editors and scriptwriters, and dramatically faster than internal production.