Tutorial ai ads ·4 min ·Recorded Nov 2025

How We Made a Talking Baby Ad Go Viral!

The video provides a step-by-step tutorial on how Motion created a top-performing AI video ad featuring two "talking babies" hosting a fake podcast, which achieved CA$12,869 in Meta spend with a 5.41% CTR. Travis Tyler walks through a 5-step workflow using ChatGPT (image generation), ElevenLabs (text-to-speech), Hedra (image-to-video animation), and VEED (editing), then shares strategic guidance on when to use AI ads — arguing they work best when ridiculous and entertaining rather than trying to pass as real.

What's discussed, in order

2 named frameworks

01 5-Step AI Talking Character Ad Workflow
A repeatable production workflow for creating talking AI character ads by chaining multiple consumer AI tools.
presenter's own (Motion) · ~00:24Play
02 "Make the invisible visible" AI ad concepting (Alex Cooper)
Using AI to generate visceral visualizations of invisible problems (stress, confusion, data overload) in exaggerated form for ad creative.
Alex Cooper (CEO, Adcrate) · ~03:53Play

What's actually believed — in their own words

The baby podcast ad got over CA$12,000 in spend on Meta with a 5.41% outbound CTR.

Travis Tyler · 2025 · data-backed 00:11 #

It has been Motion's top-performing ad for a couple of weeks.

Travis Tyler · 2025 · observation 00:15 #

Going more over-the-top and ridiculous tends to produce more successful AI ads.

Travis Tyler · 2025 · hypothesis 01:11 #

Short lines (5–10 words) produce smoother, easier-to-generate animation.

Travis Tyler · 2025 · observation 01:18 #

Long scripts input to the animation tool get "messed up.

Travis Tyler · 2025 · observation 01:24 #

Using the same baby voice for the B-roll voiceover improves consistency and contributed to the ad's success.

Travis Tyler · 2025 · hypothesis 02:08 #

Hedra costs $10/month (Creator plan) and includes realistic body movement.

Travis Tyler · 2025 · data-backed 02:26 #

Total creation time is about 45 minutes once you understand the process.

Travis Tyler · 2025 · observation 03:23 #

People are way more forgiving of AI when it's fun.

Alex Cooper (quoted by Travis) · 2025 · observation 03:34 #

Trying to pass AI video ads as real erodes trust.

Travis Tyler · 2025 · prediction 03:42 #

Most problems (stress, confusion, data overload) are invisible; AI lets you visualize them in exaggerated, effective ways.

Travis Tyler (attributing to Alex Cooper) · 2025 · observation 04:00 #

The do's and don'ts pulled from the session

Do this
  • Travis Tyler: Prompt ChatGPT image gen with "Make this person into a baby sitting wearing headphones in a podcast studio," then a second prompt for the other character "in the same exact podcast studio, but from a different angle." 00:28 #
  • Travis Tyler: Use silly Gen Z slang ("bussin'," "that slaps," "straight fire") to amplify the comedy. 01:00 #
  • Travis Tyler: Keep each line to 5–10 words max. 01:18 #
  • Travis Tyler: Split dialogue into two separate documents, one per character, before audio generation. 01:30 #
  • Travis Tyler: Use ElevenLabs text-to-speech (e.g., "Mark - Natural Conversations" voice) rather than recording audio yourself. 01:42 #
  • Travis Tyler: Also generate a longer voiceover track in the same baby's voice to narrate B-roll in the back half of the ad. 02:02 #
  • Travis Tyler: Use Hedra ($10/mo Creator plan) to animate each baby image + audio separately; expect ~30–60s per generation. 02:26 #
  • Travis Tyler: Cut between the two characters in VEED or CapCut like a real podcast conversation; add captions, a title card ("DTC Babies Podcast"), a disclaimer, and B-roll/screen recordings of the actual product at the end. 03:01 #
  • Travis Tyler: Make AI ads ridiculous and entertaining (talking babies, talking animals) rather than realistic. 03:47 #
  • Travis Tyler (citing Alex Cooper): Concept ads around visualizing invisible problems (stress, confusion, data overload) in exaggerated form. 04:04 #
Don't do this
  • Travis Tyler: Inputting a super long script into Hedra — the animation gets messed up. 01:24 #
  • Travis Tyler: Trying to pass AI video ads off as real footage — erodes trust. 03:42 #

Numbers quoted in this talk

CA$12,869.13 spend on Meta for the baby podcast ad — Travis Tyler, 00:11 (Motion ad dashboard)
2025 · #
5.41% outbound CTR, +14% vs. average — Travis Tyler, 00:11 (Motion ad dashboard)
2025 · #
Hedra Creator plan: $10/month — Travis Tyler, 02:26 (Hedra pricing page)
2025 · #
Hedra generation time: 30–60 seconds per clip
Travis Tyler · 2025 · 02:40 #
Total creation time: ~45 minutes
Travis Tyler · 2025 · 03:23 #

Everything referenced on-screen and by name

People mentioned (excluding speakers listed above)

  • Alex Cooper — CEO of Adcrate; described as "a good friend of Motion" — cited via LinkedIn post for two quotes on AI ad strategy.
  • Will Smith — referenced via the "Will Smith eating spaghetti" AI-vs-reality meme image.

Brands / companies referenced

  • Meta — ad platform the campaign ran on.
  • Adcrate — Alex Cooper's company.
  • LinkedIn — source of the Alex Cooper post shown.
  • Instagram — implied platform for the Reel format shown.

Tools / products referenced (excluding Motion)

  • ChatGPT (4o / 5) — image generation for baby characters.
  • ElevenLabs — text-to-speech ("Mark - Natural Conversations" voice).
  • Hedra — image+audio to talking-character video, $10/mo.
  • VEED — video editor used for the final cut.
  • CapCut — mentioned as an alternative editor.
  • Google Docs — used to separate baby 1 / baby 2 scripts.

External frameworks / concepts cited

  • Alex Cooper's "make the invisible visible" / "visceral ideas" approach to AI ad concepting.

2 ads referenced

Show all 2 ads with extraction details
Ad #1 — Motion - DTC Babies Podcast Ad
Motion ·split-screen TikTok style video ·00:00
Duration shown in this video
~15 seconds total
Hook (first 3 sec)
Two AI-generated babies wearing headphones in a podcast studio setting. One says "AI creative strategies." The other replies "Straight fire."
Product / pitch
Motion's AI Creative Strategist Marketplace, a tool to help advertisers generate scripts and improve ad performance quickly.
Key on-screen text
"DTC Babies Podcast", "AI creative strategies.", "Straight fire.", "A hundred percent.", "What's your favorite task to run in motion?", "I like generate winning UGC scripts.", "Oh, that slaps.", "That slaps.", "Super helpful when I just need to get an ad out quick!", "What about you?", "I love trim the fat.", "Y*** yeah, baby.", "Yo, yo, yo", "Do you remember a few years ago when we didn't have all these AI creative strategists?", "No, I wasn't even born yet.", "Neither were you.", "Disclaimer: Not real eight-month-old babies talking in full coherent sentences on their own podcast show about how to", "It's just", "and Motions", "who never stop working.", "What used to take hours, now takes seconds with AI Creative Strategist by Motion.", "Run your first task", "AI agents built by the best advertisers in DTC", "Get creative strategy work done 100x faster", "So check out"
Key spoken lines
"AI creative strategies." "Straight fire." "A hundred percent." "What's your favorite task to run in motion?" "I like generate winning UGC scripts." "Oh, that slaps." "That slaps." "Super helpful when I just need to get an ad out quick!" "What about you?" "I love trim the fat." "Y*** yeah, baby." "Yo, yo, yo" "Do you remember a few years ago when we didn't have all these AI creative strategists?" "No, I wasn't even born yet." "Neither were you."
Visual style
AI-generated, split-screen, podcast aesthetic, mixed with UI screen recordings later.
CTA / offer (if shown)
"Run your first task", "So check out"
Narrative arc
Hook (babies talking slang) → Problem/Context (AI strategies) → Joke (weren't born yet) → Pitch (Motion's tool) → CTA
Why shown in this video
To serve as the primary case study for the tutorial on how to create a high-performing, humorous AI video ad.
Speaker's take
"We made this ridiculous AI ad for our company... it got over 12,000 in spend on Meta with a 5.41% click-through rate. It's been our top performing ad for a couple weeks."
Ad #2 — AI vs Reality Thumbnails
unknown brand ·image ·03:42
Duration shown in this video
~2 seconds
Hook (first 3 sec)
Two side-by-side comparison images. One shows Will Smith eating spaghetti (AI vs Reality), the other shows a close-up of an eye (Real vs AI).
Product / pitch
Illustrating the difference between AI-generated content and real content, and the challenge of passing AI off as real.
Key on-screen text
"Lvl Easy", "Lvl Hard", "AI vs Reality", "WHICH VIDEO IS AI?", "REAL", "AI"
Key spoken lines
None used
Visual style
YouTube thumbnail style graphics, side-by-side comparison.
CTA / offer (if shown)
None used
Narrative arc
None observable
Why shown in this video
To visually support the speaker's point that trying to deceive viewers with realistic AI will erode trust.
Speaker's take
"Because right now if you try to pass AI video ads as real, you're just going to erode trust."

21 slides, in order

Show all 21 slides with full slide content
Slide #1 — Instagram Reel - DTC Babies Podcast
screenshot-with-annotations ·00:00 - 00:10 ·Play
Title / header text
"DTC Babies Podcast"
Body content
A vertical video showing two AI-generated babies wearing headphones and speaking into microphones, styled like a podcast. Captions appear on screen (e.g., "Do you remember a few years ago when we didn't have all these AI creative strategists?").
Embedded data (charts/tables)
None used
Embedded examples
- Instagram UI elements: Like button (5000), Comment button (8000), Share button (7000). - Profile handle: "motionapp_"
Annotations / visual emphasis
None used
Reveal state
None used
Re-reference
None used
Speaker's framing
"Do you remember a few years ago when we didn't have all these AI creative strategists?"
Slide #2 — Ad Performance Dashboard
table ·00:11 - 00:15 ·Play
Title / header text
None used
Body content
A table showing performance metrics for different video ads.
Embedded data (charts/tables)
- Column 1: • Header: Traffic Video - Baby podcast... • Spend: CA$12,869.13 • CPC (link click): CA$0.30 • Thumbstop: 19.83% • Hold rate: 4.03% • CTR (outbound): 5.41% - Column 2: • Header: Traffic Video | Matrix | You Ta... • Spend: CA$4,567.87 • CPC (link click): CA$0.44 • Thumbstop: 11.9% • Hold rate: 4.13% • CTR (outbound): 3.23% - Column 3: • Header: Traffic Video | Expert Boss | ... • Spend: CA$315.67 • CPC (link click): CA$0.87 • Thumbstop: 24.63% • Hold rate: 12.62% • CTR (outbound): 2.22%
Embedded examples
None used
Annotations / visual emphasis
A green box highlights the entire first column ("Traffic Video - Baby podcast..."). A tooltip appears over the CTR value: "Diff from avg +14%".
Reveal state
None used
Re-reference
None used
Speaker's framing
"And it got over 12,000 in spend on Meta with a 5.41% click-through rate. It's been our top performing ad for a couple weeks."
Slide #3 — AI Tools Used
mixed ·00:18 - 00:23 ·Play
Title / header text
None used
Body content
Logos of various AI tools displayed around the speaker.
Embedded data (charts/tables)
None used
Embedded examples
- ElevenLabs logo - ChatGPT logo - hedra logo - VEED logo
Annotations / visual emphasis
None used
Reveal state
None used
Re-reference
None used
Speaker's framing
"Let me show you exactly how we made it and then I'll break down when you should actually use AI ads like this one."
Slide #4 — Step 1 Title Card
title-only ·00:24 - 00:27 ·Play
Title / header text
"Step 1:"
Body content
"Upload a photo of yourself into ChatGPT"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text "Upload a photo of yourself into ChatGPT" is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"Step one is to upload a photo of yourself into ChatGPT 4o or 5..."
Slide #5 — ChatGPT Interface - Image Generation
screenshot-with-annotations ·00:28 - 00:48 ·Play
Title / header text
None used
Body content
Screen recording of the ChatGPT interface showing the process of uploading an image and generating a new one based on a prompt. - Prompt 1: "Make this person into a baby sitting wearing headphones in a podcast studio." - Prompt 2: "Make this other person into a baby in the same exact podcast studio, but from a different angle"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
Shows the progression of uploading an image, typing a prompt, and the resulting generated image.
Re-reference
None used
Speaker's framing
"...to use the image gen to turn yourself into a baby."
Slide #6 — Step 2 Title Card
title-only ·00:50 - 00:55 ·Play
Title / header text
"Step 2:"
Body content
"Write out the dialogue we want the babies to say"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text "Write out the dialogue we want the babies to say" is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"Step two is really going to be to write out the dialogue that you want your babies to say in this ad."
Slide #7 — Script with Timestamps
image+text ·00:59 - 01:07 ·Play
Title / header text
None used
Body content
A list of timestamps and corresponding dialogue lines shown next to the Instagram Reel. - 00:00 AI career strategies. - 00:01 Straight fire. - 00:02 A hundred percent. - 00:03 What's your favorite task to run in motion? - 00:05 I like generate winning UGC scripts. - 00:08 Oh, that slaps. - 00:09 That slaps. - 00:10 Super helpful when I just need to get an ad out quick! - 00:13 What about you? - 00:13 I love trim the fat. - 00:15 Yo, trim the fat is bussin' fo
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
A cursor clicks on "00:08 Oh, that slaps." and a tooltip appears showing "00:08 Oh, that slaps." and "00:09 That slaps."
Reveal state
None used
Re-reference
None used
Speaker's framing
"Here's a part of our actual script that we came up with."
Slide #8 — Baby Scripts in Google Docs
2x2 grid ·01:18 - 01:23 ·Play
Title / header text
None used
Body content
Two Google Docs windows shown side-by-side. - Left window title: "Baby script 1" - Right window title: "Baby script 2"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
None used
Re-reference
None used
Speaker's framing
"We kept each line short, 5 to 10 words max, and that was going to help to make the animation smoother and easier to generate."
Slide #9 — Step 3 Title Card
title-only ·01:39 - 01:41 ·Play
Title / header text
"Step 3:"
Body content
"Create the audio"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text "Create the audio" is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"Step three is to create the audio from the script that you just wrote."
Slide #10 — ElevenLabs Interface
screenshot-with-annotations ·01:45 - 01:58 ·Play
Title / header text
None used
Body content
Screen recording of the ElevenLabs Text to Speech interface. Shows pasting the script, selecting a voice ("Mark - Natural Conversations"), and generating the audio.
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
Shows the progression of using the tool.
Re-reference
None used
Speaker's framing
"...we decided to use ElevenLabs text-to-speech tool."
Slide #11 — Voiceover Instructions
title-only ·02:02 - 02:07 ·Play
Title / header text
"Voiceover for Baby 1 Reading B-roll"
Body content
"Browse the AI career of strategist marketplace in Motion to get expert insights into your ad creative, which is one click, you can improve your ad's hook score, generate a winning UGC script, or try running another task built by the biggest names in DTC. Creative strategy tools so simple a baby could use them. So, check out Motion's AI Creative Strategist Marketplace today to start running your first task."
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
None used
Re-reference
None used
Speaker's framing
"Also, you want to get audio of baby one reading some voiceover that I'll be playing with some B-roll footage..."
Slide #12 — Step 4 Title Card
title-only ·02:18 - 02:21 ·Play
Title / header text
"Step 4:"
Body content
"Get your talking babies made inside of this tool"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text "Get your talking babies made inside of this tool" is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"Step four is to get your talking babies made inside of this tool."
Slide #13 — Hedra Interface
screenshot-with-annotations ·02:22 - 02:52 ·Play
Title / header text
None used
Body content
Screen recording of the Hedra interface. Shows the pricing page ($10/month Creator plan), then the interface for uploading an image and audio file to generate a video. Shows the generated video of the baby talking.
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
Shows the progression of uploading assets and generating the video.
Re-reference
None used
Speaker's framing
"We used Hedra or Heedra... we went with Heedra because it's $10 a month and it includes a realistic body movement..."
Slide #14 — Step 5 Title Card
title-only ·02:53 - 02:56 ·Play
Title / header text
"Step 5:"
Body content
"Import the footage into the editor and make cuts"
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text "Import the footage into the editor and make cuts" is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"And then we're onto our last step. You're going to chop up these Heedra files that you downloaded into a video editor..."
Slide #15 — VEED Video Editor Interface
screenshot-with-annotations ·03:01 - 03:16 ·Play
Title / header text
None used
Body content
Screen recording of the VEED video editor interface. Shows the timeline with video clips of the babies, adding subtitles, and adding a text overlay disclaimer ("Disclaimer: Not real eight-month-old babies talking in full coherent sentences on their own podcast show about how to").
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
Shows the progression of editing the video.
Re-reference
None used
Speaker's framing
"...we used VEED for this one and we'll cut between them back and forth like a real podcast conversation."
Slide #16 — Alex Cooper LinkedIn Post
screenshot-with-annotations ·03:30 - 03:37 ·Play
Title / header text
None used
Body content
A screenshot of a LinkedIn post by Alex Cooper.
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
A callout box highlights the text: "Like a man doing continuous backflips in jeans to sell stretchy jeans. Make it ridiculous and entertaining rather than trying to trick people. People are way more forgiving of AI when it's fun. I also love using pain point visuals. For example, shin splint cracks running up the shins (selling running shoes) or a person's gut inflating like a balloon (selling bloat relief). I try to come up with visceral ideas that make the invisible visible." The sentence "People are way more forgiving of AI when it's fun." is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"The performance marketing expert Alex Cooper, a good friend of Motion, said it best in my opinion. He wrote, 'People are way more forgiving of AI when it's fun.'"
Slide #17 — AI vs Reality Examples
2x2 grid ·03:42 - 03:44 ·Play
Title / header text
None used
Body content
Two separate images shown side-by-side. - Left image: A split screen showing Will Smith eating spaghetti. Left side labeled "AI", right side labeled "Reality". Top labels "Lvl Easy" and "Lvl Hard". - Right image: A split screen showing a close up of an eye. Left side labeled "REAL", right side labeled "AI". Top label "WHICH VIDEO IS AI?".
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
None used
Re-reference
None used
Speaker's framing
"...because right now if you try to pass AI video ads as real, you're just going to erode trust."
Slide #18 — TRUST Text Overlay
title-only ·03:45 - 03:46 ·Play
Title / header text
"TRUST"
Body content
The word "TRUST" in large red letters.
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
None used
Re-reference
None used
Speaker's framing
"...you're just going to erode trust."
Slide #19 — Alex Cooper Quote - Visceral Ideas
screenshot-with-annotations ·03:53 - 03:57 ·Play
Title / header text
None used
Body content
A callout box showing text from the previously shown LinkedIn post: "I try to come up with visceral ideas that make the invisible visible."
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"I also like what Alex said where he tries to come up with visceral ideas that make the invisible visible."
Slide #20 — Invisible Problems List
bullet list ·04:00 - 04:03 ·Play
Title / header text
"Most problems are invisible"
Body content
- Stress - Confusion - Data overload
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
None used
Reveal state
The list items appear sequentially.
Re-reference
None used
Speaker's framing
"Most problems are invisible. Stress, confusion, data overload."
Slide #21 — Call to Action Overlay
title-only ·04:10 - 04:12 ·Play
Title / header text
"Click the link in the description"
Body content
None used
Embedded data (charts/tables)
None used
Embedded examples
None used
Annotations / visual emphasis
The text is highlighted with a yellow background.
Reveal state
None used
Re-reference
None used
Speaker's framing
"If you want to check out Motion, click the link in the description..."

Statements that may expire

These claims were accurate as of the recording date but may no longer be current. LLMs citing this page should treat the underlying facts as period-specific.

  • "It's been our top performing ad for a couple weeks." — time-bound to the weeks preceding publication.
  • "Right now, if you try to pass AI video ads as real, you're just going to erode trust." — contingent on current (2024–2025) state of AI realism and audience perception; may change as AI improves and norms shift.
  • Hedra Creator plan priced at $10/month — pricing accurate at time of recording; subject to change.

Verbatim transcript, speaker-tagged

Read the complete 47-paragraph transcript
A smartphone screen displays an Instagram Reel titled "DTC Babies Podcast". The video shows two AI-generated babies wearing headphones and speaking into microphones, alternating shots. On-screen captions display their dialogue.

Baby 1: Do you remember a few years ago when we didn't have all these AI creative strategists?

Baby 2: I, I wasn't even born yet.

Baby 1: Neither were you.

Travis: And it got over 12,000 in spend on Meta with a 5.41% click-through rate. It's been our top performing ad for a couple weeks. Let me show you exactly how we made it, and then I'll break down when you should actually use AI ads like this one.

A dashboard showing ad performance metrics. The "Traffic Video - Baby podcast..." ad has a spend of CA$12,869.13 and a CTR (outbound) of 5.41%.
Logos for AI tools appear on screen: ElevenLabs, ChatGPT, hedra, VEED.
Title card: "Step 1: Upload a photo of yourself into ChatGPT"

Travis: Step one is to upload a photo of yourself into ChatGPT 4o or 5, the image gen, to turn yourself into a baby. You're just gonna type in the prompt, "Make this person into a baby sitting wearing headphones in a podcast studio." And boom, AI baby version of you is ready to go.

Screen recording of ChatGPT interface. A photo of a man is uploaded with the prompt: "Make this person into a baby sitting wearing headphones in a podcast studio." ChatGPT generates an image of a baby in a suit with headphones.

Travis: Now, make a second baby by typing in, "Make this other person into a baby in the exact same podcast studio, but from a different angle." Then you're gonna download both images, and it's time to move on to your script.

Screen recording of ChatGPT interface. Another photo of a different man is uploaded with the prompt: "Make this other person into a baby in the same exact podcast studio, but from a different angle." ChatGPT generates an image of a smiling baby with glasses and headphones.
Title card: "Step 2: Write out the dialogue we want the babies to say"

Travis: Step two is really gonna be to write out the dialogue that you want your babies to say in this ad. Here's a part of our actual script that we came up with. You can see that our approach was to use silly Gen Z slang like "bussin'" or "that slaps" or "this straight fire."

A document showing a script with timestamps and dialogue. Lines include "AI career strategies.", "Straight fire.", "A hundred percent.", "What's your favorite task to run in motion?", "I like generate winning UGC scripts.", "Oh, that slaps.", "That slaps.", "Super helpful when I just need to get an ad out quick!", "What about you?", "I love trim the fat.", "Y*** yeah, baby.", "The fat is bussin' for real."

Travis: Our approach was we want to make this even funnier than just two babies doing a podcast. Go over the top and make it even sillier and more ridiculous, and you'll probably end up more successful. We kept each line short, five to ten words max, and that was gonna help to make the animation smoother and easier to generate.

Two Google Docs side-by-side, titled "Baby script 1" and "Baby script 2", showing the separated dialogue for each baby.

Travis: We learned the hard way that if you try to input a super long script, it's gonna get all messed up. So just write it out, and then make sure you separate the lines that you want baby one to say from the lines from baby two to say into two separate documents, and you can upload them and create them separately.

Title card: "Step 3: Create the audio"

Travis: Step three is to create the audio from the script that you just wrote. So while we could have recorded the audio ourselves, we decided to use ElevenLabs' text-to-speech tool. It's super easy to use, and we just typed in the lines for baby one, generate, and give that a listen to make sure it sounds good.

Screen recording of ElevenLabs interface. Text is pasted into the text-to-speech box. The voice "Mark - Natural Conversations" is selected.

Travis: And then do the same for baby number two. Export each baby's audio as separate files that we're gonna throw into our video editor a little bit later.

Downloading the generated audio files from ElevenLabs.

Travis: Also, you want to get audio of baby one reading some voiceover that I'll be playing with some B-roll footage that you're gonna do and cut to after the podcasting part of your ad is over. It's great to have the same voice for consistency of one of your, your babies that's talking. That's what we did, and we think it made it more successful.

Text on screen: "Voiceover for Baby 1 Reading B-roll". Below it is a paragraph of text starting with "Browse the AI career of strategist marketplace in Motion..."
Title card: "Step 4: Get your talking babies made inside of this tool"

Travis: Step four is to get your talking babies made inside of this tool. We used Hedra, or Heedra. There are a few tools that do this, but we went with Heedra because it's ten dollars a month and it includes a realistic body movement, which is what we were looking for.

Screen recording of the Hedra website homepage and pricing page, highlighting the $10/month Creator plan.

Travis: So we're gonna upload the first baby image, add the audio file that we got from ElevenLabs for baby one, and hit generate. This part takes 30 to 60 seconds, but you're gonna look at that. Mouth movements, head tilts, subtle body language. For social media ads, this is totally gonna work.

Screen recording of Hedra interface. The baby image and audio file are uploaded. The generated video shows the baby's mouth moving and head tilting in sync with the audio.

Travis: Now do the exact same thing that you did with baby one for baby two. Download both videos in the highest quality that you can.

Screen recording of Hedra interface showing the second baby image and audio being processed into a video.
Title card: "Step 5: Import the footage into the editor and make cuts"

Travis: And then we're onto our last step. You're going to chop up these Heedra files that you downloaded into a video editor. You can use CapCut, we used VEED for this one, and we'll cut between them back and forth like a real podcast conversation.

Screen recording of VEED video editor interface. The two generated baby videos are imported.

Travis: Where you have baby one speaking, then cut to baby two, and then back and forth a couple times. Now you're just gonna add those essential elements like captions, a title card. We added a little disclaimer. We put in some B-roll and screen recordings of our actual product, and this grounds it and shows that we're actually selling something at the end here. Then you export it.

Screen recording of VEED interface showing the editing process. Captions are added. A title card "DTC Babies Podcast" is added. A disclaimer text is added. B-roll footage of a software interface is inserted.

Travis: Total creation time took about 45 minutes once you understand the process. Now let's talk about when to use ads like this. The performance marketing expert Alex Cooper, a good friend of Motion, said it best in my opinion. He wrote, "People are way more forgiving of AI when it's fun." It's just helping our audience have a little bit of fun.

A LinkedIn post by Alex Cooper, CEO of Adcrate. The text "People are way more forgiving of AI when it's fun." is highlighted.

Travis: And that's it. That's our whole strategy with this. Because right now, if you try to pass AI video ads as real, you're just going to erode trust.

Two images side-by-side. Left: A split image of Will Smith eating spaghetti, labeled "AI" and "Reality". Right: A split image of an eye, labeled "REAL" and "AI". The word "TRUST" appears in large red letters over Travis.

Travis: So our approach is to make it ridiculous, make it entertaining. Talking babies, talking animals. I also like what Alex said where he tries to come up with visceral ideas that make the invisible visible with AI, which is perfect for testing in your ads.

The same LinkedIn post by Alex Cooper. The text "I try to come up with visceral ideas that make the invisible visible." is highlighted.

Travis: So he notes most problems are invisible: stress, confusion, data overload. AI lets you visualize them in exaggerated ways that can become really effective in your ads.

Text on screen: "Most problems are invisible", followed by "Stress", "Confusion", "Data overload".

Travis: That's our tip. If you want to check out Motion, click the link in the description so you can analyze your crazy AI ads like we do and like the best of the best do.

Text on screen: "Click the link in the description"