<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Ellis Crosby on Medium]]></title>
        <description><![CDATA[Stories by Ellis Crosby on Medium]]></description>
        <link>https://medium.com/@elliscrosby?source=rss-455933b27ced------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*bsUVaCuwRt1xY3EKpABZ5A.jpeg</url>
            <title>Stories by Ellis Crosby on Medium</title>
            <link>https://medium.com/@elliscrosby?source=rss-455933b27ced------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 20:32:03 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@elliscrosby/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Convert Youtube Video to an Article: Content Writing For People Who Hate Writing Content]]></title>
            <link>https://medium.com/@elliscrosby/convert-youtube-video-to-an-article-content-writing-for-people-who-hate-writing-content-65002db16a95?source=rss-455933b27ced------2</link>
            <guid isPermaLink="false">https://medium.com/p/65002db16a95</guid>
            <dc:creator><![CDATA[Ellis Crosby]]></dc:creator>
            <pubDate>Tue, 13 Feb 2024 16:35:31 GMT</pubDate>
            <atom:updated>2024-02-13T16:35:31.697Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Alert: This is a technical guide and requires a basic understanding of python</em></p><p>I absolutely loathe writing content. It feels cumbersome and counterintuitive to me. My wife suggests it might be because I lack a “voice” in my head, but who’s to say? The only time words flow freely is when I unleash a rant on my keyboard, much like with this piece! Hence, I frequently lean on AI to aid in content creation. Today, I’m sharing a novel method I’ve been experimenting with. Let’s dive in!</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ohhi7hhnM116DhPO.png" /></figure><p>I use sketches (hat tip Excalidraw) a lot — for everything from backend development flows, frontend UIs, in explaining ideas to people while teaching and even making nice graphs to tweet.</p><p>Recently I produced a video narrating a sales process sketch, discussing efficiency improvements via recent AI tool developments. Theoretically, providing GPT-4 with the transcript and sketch should suffice for article generation.</p><p>You can check out the the <a href="https://www.youtube.com/watch?v=dwwHe055q2E">video</a> and <a href="https://link.excalidraw.com/l/8ZlqeOQUZQK/7o24aVzar6z">sketches</a> to see what I mean.</p><p>Let’s delve into the coding aspect:</p><p>If you want to skip ahead and run this yourself you can <a href="https://colab.research.google.com/drive/1jF4znEs_tVHcW0qZ3-ozIl2ikag27yFl?usp=sharing">access the Colab notebook here</a></p><h3>Initial Setup</h3><p>Firstly, there’s a couple of packages we’ll use:</p><p>Ensure you have an OpenAI API key, available <a href="https://platform.openai.com/docs/quickstart?context=python">here</a> if you don’t already have it):</p><pre>from getpass import getpass<br><br># OpenAI API Key<br>openai_api_key = getpass(&#39;Enter your OpenAI API Key: &#39;)</pre><p>And something you should always do when playing with the AI is keeping track of the cost to avoid bankrupting yourself. This little class can handle that:</p><pre>class CostTracker:<br>    def __init__(self):<br>        self.total_cost = 0<br><br>    def track_cost(self, input_tokens, output_tokens):<br>        input_token_cost_per_mille = 0.01<br>        output_tokens_cost_per_mille = 0.03<br>        cost_to_add = ((input_tokens / 1000) * input_token_cost_per_mille) + ((output_tokens / 1000) * output_tokens_cost_per_mille)<br>        self.total_cost += cost_to_add<br>        print(f&quot;This call cost {round(cost_to_add,4)}, the total cost is now {round(self.total_cost,4)}&quot;)<br><br>    def track_image_cost(self):<br>        image_cost = 0.04<br>        self.total_cost += image_cost<br>        print(f&quot;This call cost {round(image_cost,4)}, the total cost is now {round(self.total_cost,4)}&quot;)<br><br>cost_tracker = CostTracker()</pre><h3>Getting a youtube video transcript “with AI”</h3><p>Sure, we could use Whisper or any other speech-to-text models, but that would have some cost in terms of time and money. YouTube has usually already handled this, and this Python package can extract it into a JSON. This code grabs that transcript and merges it into a single string (and yes I say “Um” a lot).</p><pre>from youtube_transcript_api import YouTubeTranscriptApi<br><br>video_link = &quot;https://www.youtube.com/watch?v=dwwHe055q2E&quot;<br>video_id = video_link.split(&quot;v=&quot;)[1]<br><br>transcript = YouTubeTranscriptApi.get_transcript(video_id)<br><br>full_transcript = &#39; &#39;.join([item[&#39;text&#39;] for item in transcript])<br>full_transcript[:100]<br><br>&gt; &quot;hey I&#39;m Ellis the lead AI developer at incremento um and last week we released an article about um a&quot;</pre><h3>Describing the image with AI</h3><p>Next, I downloaded all 4 sketches I made. We can use the <em>GPT-4-Vision</em> model to describe these sketches. The model can handle multiple images, although the OpenAI library only works with image urls so we’ll have to make an old-school request to give it our base64 images:</p><pre>def upload_files():<br>  from google.colab import files<br>  uploaded = files.upload()<br>  for k, v in uploaded.items():<br>    open(k, &#39;wb&#39;).write(v)<br>  return list(uploaded.keys())<br><br>uploaded_screenshots = upload_files()</pre><pre>import base64<br>import requests<br><br><br># Function to encode the image<br>def encode_image(image_path):<br>  with open(image_path, &quot;rb&quot;) as image_file:<br>    return base64.b64encode(image_file.read()).decode(&#39;utf-8&#39;)<br><br># Getting the base64 string<br>base64_images = [encode_image(image_path) for image_path in uploaded_screenshots]<br>message = &quot;I have sketched out a few things here to illustrate how my company is augmenting sales teams. There are 4 sketches, one of the sales process and 3 of the various tools we have built to augment parts of this process. The first is a tool that extracts insights from Slack conversations, and can act as a chatbot to respons back with information. The second is a chatbot that checks a leads qualification status before they pass to a salesperson - the image shows two conversations, one qualified and one not, the unqualified one is directed to a 2nd chatbot that sells them a cheaper course. The last is a tool that enriches leads intelligently by reasoning through the information missing and how it can find it. Please describe the images and processes for the three tools in as much detail as possible so that we can write some content about them.&quot;<br><br>headers = {<br>  &quot;Content-Type&quot;: &quot;application/json&quot;,<br>  &quot;Authorization&quot;: f&quot;Bearer {openai_api_key}&quot;<br>}<br><br>content = [<br>        {<br>          &quot;type&quot;: &quot;text&quot;,<br>          &quot;text&quot;: message<br>        },<br>      ]<br><br>for base64_image in base64_images:<br>  content.append(<br>        {<br>          &quot;type&quot;: &quot;image_url&quot;,<br>          &quot;image_url&quot;: {<br>            &quot;url&quot;: f&quot;data:image/jpeg;base64,{base64_image}&quot;<br>          }<br>        })<br><br>payload = {<br>  &quot;model&quot;: &quot;gpt-4-vision-preview&quot;,<br>  &quot;messages&quot;: [<br>    {<br>      &quot;role&quot;: &quot;user&quot;,<br>      &quot;content&quot;: content<br>    }<br>  ],<br>  &quot;max_tokens&quot;: 3000<br>}<br><br>response = requests.post(&quot;https://api.openai.com/v1/chat/completions&quot;, headers=headers, json=payload)<br><br>image_descriptions = response.json()[&#39;choices&#39;][0][&#39;message&#39;][&#39;content&#39;]<br><br># Don&#39;t forget to track the cost!<br>input_tokens = response.json()[&#39;usage&#39;][&#39;prompt_tokens&#39;]<br>output_tokens = response.json()[&#39;usage&#39;][&#39;completion_tokens&#39;]<br>cost_tracker.track_cost(input_tokens,output_tokens)<br><br>image_descriptions[:300]<br><br>&gt; &quot;&quot;&quot;Certainly, let&#39;s go through each image and describe the processes and tools illustrated.<br><br>The first image appears to be a detailed sketch of an interactive and responsive sales tool that&#39;s based on extracting insights from Slack conversations. The sketch portrays a conversation between two individua&quot;&quot;&quot;</pre><h3>Converting your script into text, article ideas &amp; structures with AI</h3><p>Next we’ll start putting the article together. A best practice is to separate the writing process into 1) ideas 2) structure and 3) the writing. This will give you more interesting ideas, a more coherent structure and a longer article than if the article is being generated with less guidance. This also means we can have some human feedback in the loop on which idea is more interesting. Here I’ve added the instruction in the system prompt and the instruction subject to the message. I’ve merged the ideas and structure step here as the two can go together without a lot of additional token overhead.</p><p>There are some of the usual prompt engineering best practices here such as:</p><ul><li>Giving role context (“You are a content writer…”)</li><li>Giving task context (“For each article I’ll send you…”)</li><li>Output constraints (“using only the information from”)</li><li>Well labeled and clearly separated task context (“Description of sketches:” &amp; “\n########\n”)</li><li>Giving a goal (“To get leads from Sales Directors”)</li><li>Detailing the task last to ensure full comprehension (“Create 3 ideas for this article”)</li><li>And even the “take a deep breath” modifier that <em>sometimes</em> can improve the output</li></ul><p><em>Note that we want some creativity in this ideation step so I’ve set the temperature to 0.85. I want the model to stop writing once it has completed the task, rather than run out of tokens so I’ve set max_tokens to 3000 too</em></p><pre>from openai import OpenAI<br><br>openai_client = OpenAI(api_key=openai_api_key)<br><br>rough_idea_for_the_article = &quot;How to 10X sales for a salesteam using AI&quot;<br>goal_for_the_article = &quot;To get leads Sales Directors who are interested in us building similar solutions for them&quot;<br><br>system_prompt = &quot;You are a content writer working for Incremento. Your boss Ellis is lazy and doesn&#39;t write content by himself, instead he draws sketches and records himself talking about the content.&quot;<br>system_prompt += &quot;For each article I&#39;ll send you a description of his sketches, a transcript of his narration, the idea he has for the article, and the goal of the article, and your current task in creating the article.&quot;<br>system_prompt += &quot;Your job is to make sense of it all and write a great piece of content using only the information from the narration and the sketches. First take a deep breath, I understand that this is a frustrating job.&quot;<br><br>message=&quot;Description of sketches: \n&quot;<br>message+=image_descriptions<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Narration transcript: \n&quot;<br>message+=full_transcript<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Idea for the article: \n&quot;<br>message+=rough_idea_for_the_article<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Goal for the article: \n&quot;<br>message+=goal_for_the_article<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Current Task: \n&quot;<br>message+=&quot;Create 3 ideas for this article, including a small summary and the outline of the article. These will be sent to your boss for him to choose the best one&quot;<br>message+=&quot;\n########\n&quot;<br><br>response = openai_client.chat.completions.create(<br>    model = &quot;gpt-4-turbo-preview&quot;,<br>    messages = [<br>        {&quot;role&quot;:&quot;system&quot;,&quot;content&quot;:system_prompt},<br>        {&quot;role&quot;:&quot;user&quot;,&quot;content&quot;:message}<br>    ],<br>    temperature=0.85,<br>    max_tokens=3000<br>)<br><br>content = response.choices[0].message.content<br><br>input_tokens = response.usage.prompt_tokens<br>completion_tokens = response.usage.completion_tokens<br>cost_tracker.track_cost(input_tokens,output_tokens)<br>print(content)<br><br>&gt;&quot;### Idea 1: Transform Your Sales Process: AI-Powered Strategies for Exponential Growth<br><br>**Summary:**<br>This article will dive into the transformative power of AI in the sales process, showcasing real-world applications and client success stories from Incremento. By integrating AI tools for lead qualification, enrichment, and sales process efficiency, businesses can significantly enhance their sales output. The ultimate goal is to spark interest among Sales Directors in exploring these AI solutions to 10X their sales team&#39;s performance.<br><br>**Outline:**<br>1. **Introduction** <br>    - Brief overview of the challenges faced by sales teams<br>    - The promise of AI in revolutionizing sales processes<br>2. **Case Study: The AI-Powered Sales Qualification Chatbot**<br>    - Description of the chatbot&#39;s mechanism and its dual focus on qualifying leads and maximizing value from unqualified leads<br>    - Client success story highlighting increased sales and efficiency<br>3. **The Power of Lead Enrichment Tools**<br>    - How AI-driven lead enrichment can fill information gaps and improve lead quality<br>    - Example of a client who experienced a more refined lead pool and better-targeted sales strategies<br>4. **Streamlining Sales Operations with AI Insights**<br>    - Insight into the tool that integrates with Slack for capturing valuable sales information and tasks<br>    - A success story of how this tool facilitated better focus, follow-ups, and closed sales<br>5. **Conclusion**<br>    - Recap of the benefits of incorporating AI into the sales process<br>    - Call to action for Sales Directors to reach out for a custom AI solution consultation with Incremento<br><br>---<br>...&quot;</pre><h3>Writing a full article using AI</h3><p>The previous step gave us 3 ideas — I chose my favourite and added that back into a prompt fairly similar to the above. I’ve started this prompt from scratch rather than continuing a chat as I’ve found the accuracy in completing the task to be much higher when keeping the conversation just to the relevant prompt points, rather than things left over from earlier calls (this includes starting new chats for each message rather than continuing existing ones in the ChatGPT interface).</p><p>The task here is a pretty basic “Write an article” on with the below features:</p><ul><li>Constraint of only using the information given to keep things relevant and reduce hallucination/imagination (“using only the information given above, sticking to the idea and outline given” &amp; a temperature setting of 0.6)</li><li>Markdown formatting to work with the blog tool I use (“ Output this with markdown formatting”)</li><li>Output length controls (“This article should be detailed and verbose”, “6–12 paragraphs, totalling around 2000 words” &amp; max_tokens set to the models maximum 4096). _N.b. this isn’t likely to end up at 2000 words, but adding this wil increase it to ~800–1000 rather than 500</li><li>Weak tone controls (“casual, professional tone”). We could do a lot more here but the use case doesn’t call for a more controlled or stylised tone</li></ul><pre>idea_for_the_article = &quot;&quot;&quot;**Leveraging AI for Unprecedented Sales Growth: A Guide for Sales Directors**<br><br>*Summary*: This article is a comprehensive guide for Sales Directors on leveraging AI to overhaul their sales process, focusing on Incremento&#39;s expertise in building AI solutions that streamline lead qualification, enrich leads, and enhance sales efficiency. Through detailed explanations and client success stories, Sales Directors will be encouraged to partner with Incremento for their AI needs.<br><br>&quot;&quot;&quot;<br>outline = &quot;&quot;&quot;**Outline:**<br>1. **Introduction**<br>    - The evolving role of AI in the competitive landscape of sales<br>2. **Revolutionizing Lead Qualification with AI**<br>    - An in-depth look at AI chatbots&#39; role in identifying and nurturing qualified leads<br>    - Highlighting a transformative impact on a client&#39;s sales process<br>3. **The Strategic Advantage of AI Lead Enrichment**<br>    - Discussing how AI can uncover valuable lead insights for more effective sales pitches<br>    - Client story: achieving a more accurate and action-driven lead database<br>4. **Boosting Sales Efficiency Through AI Insights**<br>    - Examination of how AI tools can extract actionable tasks and insights from sales communications<br>    - Sharing a case where improved sales team focus led to dramatic increases in sales outcomes<br>5. **Conclusion**<br>    - Reinforcing the need for Sales Directors to adopt AI for sales growth<br>    - A call to action for consulting with Incremento to explore bespoke AI solutions&quot;&quot;&quot;<br><br>system_prompt = &quot;You are a content writer working for Incremento. Your boss Ellis is lazy and doesn&#39;t write content by himself, instead he draws sketches and records himself talking about the content.&quot;<br>system_prompt += &quot;For each article I&#39;ll send you a description of his sketches, a transcript of his narration, the idea for the article, the goal of the article, the outline for this article, and your current task in creating the article.&quot;<br>system_prompt += &quot;Your job is to make sense of it all and write a great piece of content using only the information from the narration and the sketches. First take a deep breath, I understand that this is a frustrating job.&quot;<br><br>message=&quot;Description of sketches: \n&quot;<br>message+=image_descriptions<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Narration transcript: \n&quot;<br>message+=full_transcript<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Idea for the article: \n&quot;<br>message+=idea_for_the_article<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Goal for the article: \n&quot;<br>message+=goal_for_the_article<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Outline for the article: \n&quot;<br>message+=outline<br>message+=&quot;\n########\n&quot;<br><br>message+=&quot;Current Task: \n&quot;<br>message+=&quot;Write this article using only the information given above, sticking to the idea and outline given. Output this with markdown formatting. This article should be detailed and verbose, therefore it should contain between 6-12 paragraphs, totalling around 1000 words. This should have a casual, professional tone.&quot;<br>message+=&quot;\n########\n&quot;<br><br>response = openai_client.chat.completions.create(<br>    model = &quot;gpt-4-turbo-preview&quot;,<br>    messages = [<br>        {&quot;role&quot;:&quot;system&quot;,&quot;content&quot;:system_prompt},<br>        {&quot;role&quot;:&quot;user&quot;,&quot;content&quot;:message}<br>    ],<br>    temperature=0.6,<br>    max_tokens=4096<br>)<br><br>content = response.choices[0].message.content<br><br>input_tokens = response.usage.prompt_tokens<br>completion_tokens = response.usage.completion_tokens<br>cost_tracker.track_cost(input_tokens,output_tokens)</pre><p>Woo! We now have an article. If you want to read the output in full you can do so on our blog in our <a href="https://incremen.to/ai-for-sales-prospecting">AI for sales prospecting</a> article. This has been tweaked <em>very slightly</em> in line with our SEO strategy but is basically the complete output of the AI.</p><h3>Generating images for the article with AI</h3><p>Next we’ll get an image to go alongside this on our blog. My go to strategy for blogs is 1) Get an idea from the blog with an LLM and 2) Using that idea as a prompt with an image generation model. Because of how limited our control is over LLM generated ideas/prompts, and how forgiving DALL-E 3 is with bad prompts, we’ll use that image generation model.</p><p>We can generate the image idea by passing in some relevant information on the article from above, e.g. the idea or text — I’ve gone for the full text here but if we were using this in production we’d probably want to save on tokens by just using the idea. I’ve also included: -a bit of guidance (“colourful, retro and realistic”) -a small request to fit in with the rest of our blog (“Include some robots”) -a format control, so that we can put the output straight into the image model (“Respond only with the idea”)</p><p>First the image idea:</p><pre>prompt = &quot;Think of an idea for an image for this blog post. This should be colourful, retro and realistic. It should have no text in the image. Include some robots. Respond only with the idea.&quot;<br>prompt += &quot;\n Article: \n&quot;<br>prompt += content<br><br>response = openai_client.chat.completions.create(<br>    model = &quot;gpt-4-turbo-preview&quot;,<br>    messages = [<br>        {&quot;role&quot;:&quot;user&quot;,&quot;content&quot;:prompt}<br>    ],<br>    temperature=0.85,<br>    max_tokens=4096<br>)<br><br>image_prompt = response.choices[0].message.content<br><br>input_tokens = response.usage.prompt_tokens<br>completion_tokens = response.usage.completion_tokens<br>cost_tracker.track_cost(input_tokens,output_tokens)<br>print(image_prompt)<br><br>&gt; &quot;An image featuring a vibrant, 80s-style office space with pastel and neon colors highlighting the retro aesthetic. In the center, a humanoid robot with a sleek, chrome design, reminiscent of vintage sci-fi, is seated at a desk. The desk is equipped with futuristic yet retro-looking computers and gadgets, pulsing with soft neon lights. The robot is engaging in a conversation with holographic displays that project images of potential clients, symbolizing AI&#39;s role in lead qualification and enrichment. Around the robot, other robots perform various tasks such as organizing files, analyzing data on screens, and interacting with more holograms, showcasing the efficiency and multitasking capabilities AI brings to the sales process. The background subtly features analog synthesizers and vinyl records, reinforcing the retro vibe, while the windows offer a view of a city skyline at dusk, bathed in the glow of neon signs and holographic advertisements, hinting at a world where AI seamlessly integrates into everyday business operations.&quot;</pre><p>This is quite an interesting, albeit very detailed prompt. Thankfully DALL-E 3 will automatically optimise the prompt further as it sees fit, so we should get something pretty nice.</p><p>Then we generate the image:</p><pre>response = openai_client.images.generate(<br>      model=&quot;dall-e-3&quot;,<br>      prompt=image_prompt,<br>      quality=&quot;standard&quot;,<br>      size=&#39;1024x1024&#39;,<br>      n=1<br>  )<br>image_url = response.data[0].url<br>cost_tracker.track_image_cost()</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*minoqZeoE7G3zx2Y.png" /></figure><p>Great! This is exactly the kind of thing we wanted, and it does kind of look like a sales team of robots.</p><h3>Conclusion</h3><p>So how does this article look overall? It’s decent, maybe even good. While it isn’t a masterpiece it is a decent article that accurately showcases a few of the things we’ve built, and I did save time and effort in writing it.</p><p>We could improve on this further with more control over the tone. For example we could first have the AI describe the features/tone of one of our previous blogs and have it write this article in the same style. We could have also split up the AI calls to create each secion individually and put it together at the end to ensure a longer length.</p><p>Let’s not forget the cost:</p><pre>print(f&quot;This entire article cost: ${round(cost_tracker.total_cost,4)}&quot;)<br>&gt; This entire article cost: $0.3004</pre><p>Not bad! $0.30 is pretty cheap, and a lot cheaper than the cost of my time if I’d have written this. If we really wanted to we could optimise the prompts and the chaining of these to reduce the tokens used — or optimise the prompts to work with cheaper models — but as we’re not putting this into production $0.30 is an acceptable cost.</p><p>Links:</p><p><a href="https://www.youtube.com/watch?v=dwwHe055q2E&amp;t=15s">Youtube video</a></p><p><a href="https://colab.research.google.com/drive/1jF4znEs_tVHcW0qZ3-ozIl2ikag27yFl?usp=sharing">Colab notebook</a></p><p><a href="https://incremen.to/ai-for-sales-prospecting">Finished article</a></p><p>If you want to talk to one our our friendly AI developers to see how we could help your content strategy you can book a free 15 minute consultation <a href="https://calendly.com/ellis-incremento/15min">through this link here</a>.</p><p><em>Originally published at </em><a href="https://incremen.to/youtube-video-to-text"><em>https://incremen.to</em></a>/<em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=65002db16a95" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to create a new, personalised bedtime story each day with Scarlett Panda]]></title>
            <link>https://medium.com/@elliscrosby/how-to-create-a-new-personalised-bedtime-story-each-day-with-scarlett-panda-eabcb730d517?source=rss-455933b27ced------2</link>
            <guid isPermaLink="false">https://medium.com/p/eabcb730d517</guid>
            <category><![CDATA[bedtime-stories]]></category>
            <category><![CDATA[parenting]]></category>
            <category><![CDATA[bedtime-stories-for-kids]]></category>
            <dc:creator><![CDATA[Ellis Crosby]]></dc:creator>
            <pubDate>Tue, 27 Dec 2022 13:17:27 GMT</pubDate>
            <atom:updated>2022-12-27T13:17:27.470Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*g6_dx0gGWywB9h8RVqGs8w.png" /><figcaption>A father reading to his daughters</figcaption></figure><p><em>In 2022, my wife and I had our first child (Scarlett). Since a child, I’ve always looked forward to reading bedtime stories to my own children one day; so this became part of our bedtime routing once Scarlett started sleeping through the night. We read a combination of books that we’d bought and short stories we’d found online — but this led to us repeating stories quite often or spending an annoying amount of time searching. Of course, my 6 month old baby didn’t mind but it did take away some of the magic for me. So my wife and I put our heads together and created an app that scratched our itch, as well as those of other parents we talked to. Using generative AI we built an app…</em></p><p>Scarlett Panda is a fun and interactive app that allows you to create personalised bedtime stories and illustrations for your child in just 30 seconds. With Scarlett Panda, you can add characters (your child and their friends, family and toys), settings, and plot elements to craft unique and imaginative stories that will capture your child’s attention and give the parent something new to read each night.</p><p>So, how do you get started with Scarlett Panda? Follow these simple steps:</p><ol><li>Visit the <a href="https://www.scarlettpanda.com/">Scarlett Panda</a> website and create an account. You can also try a story for free without an account <a href="https://app.scarlettpanda.com/try-a-story">here</a>.</li><li>Choose a character for your story. You can add your child or another character they’d like to see, including friends, family members or even their toys or pets. You can choose a character that your child is already familiar with, or try something new and exciting.</li><li>Select a setting for your story. Scarlett Panda offers a variety of settings, from cozy homes to magical forests and beyond. Something that is relevant to your child’s life, or something from their imagination. Choose a setting that your child will find interesting and engaging.</li><li>Choose a moral or learning for your story. Scarlett Panda can incorporate any learning into the story, whether it’s one relevant to your child’s life (e.g. “don’t be scared about going to the doctors”) or something more generic (e.g. “listen to your parents”)</li><li>Generate the story. This takes around 30 seconds so you can create it as you need, or generate it in advance and save it for later.</li><li>Read the story to your child at bedtime. Enjoy the bonding time that reading with your child creates.</li><li>Come back tomorrow evening for another!</li></ol><p>Creating a new bedtime story each day with Scarlett Panda can be a fun and rewarding activity for both you and your child. Here are a few tips and ideas to help you come up with unique and imaginative stories:</p><ul><li>Experiment with different characters and settings to keep things fresh and exciting.</li><li>Incorporate your child’s interests and hobbies into the story to make it more meaningful and engaging.</li><li>Use your child’s name and include them in the story as a character to make them feel special and involved.</li><li>Encourage your child to come up with their own ideas and plot elements to help them develop their creativity and problem-solving skills.</li></ul><p>In conclusion, creating a new bedtime story each day with <a href="https://scarlettpanda.com/">Scarlett Panda</a> is a fun and beneficial activity for children. It can improve their language skills, promote a love of reading, and encourage their creativity and imagination. So why not give it a try with your own child and see the magic unfold?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eabcb730d517" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Exploring the Benefits and Concerns of AI in Children’s Stories: A Look at Scarlett Panda’s App]]></title>
            <link>https://medium.com/@elliscrosby/exploring-the-benefits-and-concerns-of-ai-in-childrens-stories-a-look-at-scarlett-panda-s-app-3954c29635b9?source=rss-455933b27ced------2</link>
            <guid isPermaLink="false">https://medium.com/p/3954c29635b9</guid>
            <category><![CDATA[childrens-stories]]></category>
            <category><![CDATA[generative-ai]]></category>
            <category><![CDATA[gpt-3]]></category>
            <category><![CDATA[dalle-2]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Ellis Crosby]]></dc:creator>
            <pubDate>Mon, 12 Dec 2022 08:56:16 GMT</pubDate>
            <atom:updated>2022-12-12T08:56:16.305Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1014/0*sJ-sr_aVB6OMVgud" /><figcaption>A robot reads a story to a child</figcaption></figure><h3>The recent growth in the use of artificial intelligence in more consumer-facing apps has sparked a debate about the potential implications of this technology.</h3><p>My partner and I recently soft-launched <a href="https://scarlettpanda.com/">scarlettpanda.com</a>, a product that generates children’s stories using AI — with the aim of allowing parents to read fresh stories to their children each night that can be completely personalised. Some argue that relying on AI to create stories for children is dystopian, as it could lead to the loss of human creativity and the homogenisation of children’s literature. Others, however, believe that AI can be a valuable tool in the creation of children’s books, allowing for more diverse and inclusive storytelling.</p><p>As the creators of an AI-powered tool for creating personalised children’s stories, we have seen firsthand the potential benefits of using this technology in storytelling. The vast majority of feedback we’ve received has been hugely positive — parents have told us that our app has given them plenty of fresh stories to read to avoid repeating the same ones over and over, while giving their child more control over what the story of the day will be about (“can today’s story be about firemen?”).</p><p>The personalisation has had some wholesome impacts, one mother created a story for her daughters that explains their grandmother’s dementia in a kid-friendly way. A mother with a dyslexic child who was previously averse to reading used our app to get the child excited about reading by being part of the creation process. That’s on top of the fun stories that have been created about children’s favourite toys going on adventures or children overcoming the obstacles in their lives (such as going to the dentist).</p><p>Our app can also be used to create more diverse and inclusive children’s stories. By using natural language processing and other AI techniques, it is possible to generate stories that feature a wider range of characters and experiences. This can be particularly valuable in promoting representation and inclusivity in children’s literature, which has historically been dominated by a narrow range of voices and perspectives and is slow to catch up.</p><p>There are wider, ongoing conversations about the risk generative AI poses a risk to the jobs of real people and the potential for plagiarism of the work of real artists in the models used. While this article is aimed at the specific impacts on children’s stories, we remain part of those conversations. Our app is designed to work alongside, rather than replace, traditional children’s books — and our aim is not to impact the jobs currently involved in that area.</p><p>Specific to children’s stories, we’ve had some people try to paint a picture of a world in which parents are replaced by robots, but our app is aimed more at strengthening the quality time a parent has with their children by making bedtime stories fresher, more engaging and more relevant to a child. Rather than replacing other books, we’re allowing parents to add to their collection of stories with some that are more personalised for the child. We’re also creating a tool that helps the child to become more collaborative in the story that is being read, assisting creativity rather than replacing it. This is something that we keep in mind as we build our app, to keep it as something to improve the connection between a parent and child rather than replace the need for it.</p><p>Overall, the use of AI in children’s stories has the potential to be a valuable tool for promoting personalisation, diversity, and inclusivity in storytelling. However, it is important for creators of AI-powered tools for children’s literature to carefully consider the potential risks and ethical implications of using this technology. We’re excited to keep this conversation evolving and open-ended, so if you have any additional thoughts to share please <a href="https://scarlettpanda.com/">get in touch with us</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3954c29635b9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Poor account structure is a bigger problem than bad ad copy for Google Ads campaigns]]></title>
            <link>https://medium.com/@elliscrosby/poor-account-structure-is-a-bigger-problem-than-bad-ad-copy-for-google-ads-campaigns-d426e83b4250?source=rss-455933b27ced------2</link>
            <guid isPermaLink="false">https://medium.com/p/d426e83b4250</guid>
            <category><![CDATA[digital-marketing]]></category>
            <category><![CDATA[google-ads]]></category>
            <category><![CDATA[copywriting]]></category>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[ppc]]></category>
            <dc:creator><![CDATA[Ellis Crosby]]></dc:creator>
            <pubDate>Wed, 19 Jan 2022 14:37:52 GMT</pubDate>
            <atom:updated>2022-01-19T14:37:52.909Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lAvO3BnlymmND_L-UKggxA.png" /><figcaption>Adtiplier.ai — the Google Ads ad copy suite</figcaption></figure><p>3 months ago we launched the beta of Adtiplier.ai — the Google Ads ad copy suite that helps advertisers to improve CTRs and conversion rates by generating ad copy, analysing the characteristics of high performing assets, and distributing ad copy through the account. As we make the last few changes to our tool before launching in Q2, we’re reflecting on what we’ve learned about ad copy from the £2.8m of ad spend that has been run through our tool in the last 3 months.</p><p>At the start of the project we predicted that most mature Google Ads accounts would contain a mixture of high and low performing copy; our vision was that our tool would help advertisers to remove low performing copy and replace it with new variants of high performing copy. However what we learned was that over 90% of accounts analysed had more assets with insufficient data to operate correctly than they had low performing assets.</p><blockquote>Over 90% of accounts have more of an issue with insufficient data than low performing ad copy</blockquote><h3><strong>Data management is more important than copywriting</strong></h3><p>When an asset doesn’t get enough impressions, the Google Ads algorithms can’t determine how well it performs, and an ad will struggle to be optimised effectively; we’ve seen a correlation between assets with insufficient data and low performance, so this is certainly an issue. This can happen when one or both of these criteria are met, with regard to how many impressions are captured: 1) ad group structures are too granular, or 2) ads contain too many assets (headline &amp; description variants).</p><p>Both of these cases can result in impressions being spread out too thinly between assets to the point where Google Ads can’t determine if they are high or low performing, and optimise the combination of assets accordingly.</p><p>Three years ago we were huge advocates of hyper-granular, SKAG structures as a way of maximising the control an advertiser has over their account. The growing focus in Google Ads on automated solutions means that how data is managed and balanced in accounts is crucial to performance; sadly most keywords don’t get enough impressions to effectively run in their own ad group compared to grouped with others.</p><h3><strong>How much data does an ad need to run effectively?</strong></h3><p>We analysed anonymised data from our platform to find out how advertisers can ensure assets are getting enough data. We’ve found that this depends a lot on two factors: 1) the impressions an ad receives and 2) the number of assets (headlines/descriptions) are in an ad.</p><p>For ads with just 4 headline assets only 15k monthly impressions were required to be sure of reaching a sufficient data level, but the same certainty required 30k impressions for ads with 6 headlines. Ads with 10 headlines reached 80% data sufficiency at around 35k ads, but ads with 15 headlines almost never reached higher than 30% data sufficiency. Descriptions follow an even sharper trend, with only 40% of ads having sufficient data at 30k impressions and 80% at 45k (likely due to the more subtle differences that descriptions have compared to headlines)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*UCfaEWng6O9XJ72-" /><figcaption>The % of ads with sufficient data depending on the monthly impressions and number of headline assets</figcaption></figure><h3><strong>How can advertisers optimise accounts for this?</strong></h3><p>The data suggests that advertisers need to carefully balance the number of impressions that each ad receives with the number of assets included in it.</p><p>For an ad group with 15k impressions per month, it would be best to have one responsive search ad with just 4 headlines — alternatively more could be included if the ad group were to be merged with one similar to it to increase the volume of impressions captured.</p><p>Once we identified that data availability was a key issue in the performance of ad copy, we set about building features to help solve this for advertisers. We’ll soon be adding two useful features for this:</p><ul><li>Recommended ad group merges — in which we identify ad groups that aren’t reaching data sufficiency and find ad groups with similar keywords and ads to merge them into</li><li>Recommended asset reductions — for ads that have more assets than would be optimal for the impressions it receives, we recommend which assets should be removed (based on how they perform in other campaigns)</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RFOVxKjGu47byouU" /><figcaption>Adtiplier.ai features to help with insufficient ad data</figcaption></figure><h3>Join the Adtiplier.ai beta to find out if your ads have data insufficiency issues</h3><p>Adtiplier.ai is still in beta but will be launching in Q2 of 2022. If you’re interested in any of the features above, sign up at <a href="http://adtiplier.ai/">adtiplier.ai</a> to join the waiting list, or preorder today to get lifetime access at a highly discounted rate.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d426e83b4250" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>