Artificial Societies · Oct 2026
Retell AI - Introducing Tell-1, Retell’s LLM purpose-built for contact centers.
Retell AI · Nov 2026 · 3:03 · AI-annotated
- Length
- 3:03
- Scenes
- 35 · 5.2 s avg
- Product on screen
- 0:00
- Features shown
- 3
- Voiceover
- Yes
Retell AI’s 3:03 launch film runs 35 scenes, about 5.2 s each. The product is on screen from the first frame and 3 features are revealed. There is no explicit call to action; a voiceover carries the film. Founder interviews on a dark couch alternate with clean motion graphics (diagrams, waveforms, UI cards) on pale and navy gradients, and each claim is built word by word over the speaker. A steady cut rhythm, the high-contrast serif headlines, and the recurring blue sphere and dot-ring marks give the launch a polished, technical signature.
Is this your launch? to your site or press kit.
More like this
Nous Research · Oct 2024
Hopper Research · Oct 2026
Retell AI - Introducing Tell-1, Retell’s LLM purpose-built for contact centers., scene by scene
· 35 scenes as text
01 · Product intro · 0:00–0:01
A man in a black shirt sits on a dark couch before a window lined with cypress trees, gesturing with open palms as he speaks. A white name caption fades in at the lower left. The shot is a static medium frame.
On-screen text: “Zexia Zhang”, “Co-founder & CTO, Retell”
Camera: Static
02 · Product intro · 0:01–0:03
A man in a black shirt speaks in a shallow-focus interview shot, a small circular Tell-1 logo floats at right. A dark blue circle expands from the right edge, wiping over the frame while the lower-left name caption holds.
On-screen text: “Zexia Zhang”, “Co-founder & CTO, Retell”, “Tell-1”
Techniques: Masking
Camera: Static
03 · Product intro · 0:03–0:08
A glowing blue sphere logo sits at the top of a dark blue gradient. The word INTRODUCING and the serif title Tell-1 fade in from blur, then the tagline The best model for contact centers. appears word by word below. All text fades out, leaving the logo.
On-screen text: “INTRODUCING”, “Tell-1”, “The best model for contact centers.”
Techniques: Kinetic type
Camera: Static
04 · Problem · 0:08–0:12
A man in a black sweater sits on a dark sofa before floor-to-ceiling windows with cypress trees outside, hands clasped, speaking to camera. White caption lines build up on the left, one line scaling in, while the shot stays locked off.
On-screen text: “We've integrated”, “with different”, “model providers”, “and mainstream”, “LLMs for the”, “response engine.”
Techniques: Kinetic type
Camera: Static
05 · Problem · 0:12–0:15
A light gradient background holds a two-row diagram under a RESPONSE ENGINE header. The cascading row (Speech to text, LLM, Text to speech) builds left to right, the duplex row (Speech-to-speech model, LLM) appears below, then both LLM boxes fill blue.
On-screen text: “And no matter if you're using cascading”, “or duplex model, they all needed LLM”
Camera: Static
06 · Problem · 0:15–0:16
A blue two-shape logo mark sits on a pale field with faint labels and a bottom caption. A white panel wipes in from the right as the mark morphs into a gradient sphere with a thin orbit ring, and the word 'The' fades in above it.
On-screen text: “or duplex model, they all needed LLM”, “The”
Techniques: Morph, Masking
Camera: Static
07 · Problem · 0:16–0:21
Split screen: a man in a black sweater is interviewed on the left; on the right a two-tone blue sphere rests on a white grid as a particle ring swirls around its base. Captions fade in below the sphere, and the right panel narrows at the end.
On-screen text: “The LLM is the center”, “complicated tasks”, “nuanced situations”
Techniques: Split screen, Particles
Camera: Pan
08 · Problem · 0:21–0:22
A man in a black t-shirt sits on a dark couch in a softly lit room with blurred greenery behind, speaking to an off-camera interviewer. His name and title fade in at lower left as subtitles run along the bottom.
On-screen text: “Jerry Cheung”, “Senior ML Engineer, Retell”, “For the contact center setting, a lot of”
Camera: Static
09 · Problem · 0:22–0:25
A man in a black t-shirt sits on a dark couch before floor-to-ceiling windows showing trees, speaking to camera. A white name lower-third holds at left while subtitles cycle at the bottom. The shot is locked off.
On-screen text: “Jerry Cheung”, “Senior ML Engineer, Retell”, “For the contact center setting, a lot of”, “times people just add more and more”, “instructions to the system prompt to”
Camera: Static
10 · Problem · 0:25–0:32
A white System prompt card sits centered on a soft blue gradient. Instruction lines type in one by one, the card grows, then the view zooms in and the list scrolls up as one line highlights in orange with a Missed tag.
On-screen text: “instructions to the system prompt to”, “steer the model's behavior.”, “So these additional prompts make the task”, “difficult to follow instructions.”
Techniques: UI zoom
Camera: Push in
11 · Problem · 0:32–0:35
A man in a black T-shirt sits in profile on a dark couch, talking, against a soft-focus green foliage background. White text “Duplex models” appears top left, then “often” fades in beneath it, while a subtitle line runs along the bottom.
On-screen text: “In terms of the duplex models, these”, “Duplex models”, “often”, “for the harder”
Camera: Static
12 · Problem · 0:35–0:43
A man in a black t-shirt sits on a dark couch before a window of tall cypresses, talking and gesturing. Large white text builds word by word on the left, swaps to a second block, then bottom subtitles take over.
On-screen text: “Duplex models often perform delegations”, “for the harder tasks to a powerful LLM in the background.”, “For example, when chatting with other”, “duplex models, I would often hear, 'Wait'”
Techniques: Kinetic type
Camera: Static
13 · Problem · 0:43–0:46
A man in a black tee sits in profile on a dark couch against a blurred leafy window. A dark blue chat bubble labeled DUPLEX AGENT fades in at left, its white quote typing out word by word, while a white caption tracks his speech along the bottom.
On-screen text: “duplex models, I would often hear, “Wait”, “DUPLEX AGENT”, ““Wait”, “a second, let me think,””, ““Wait a second,””, “let me think...””
Techniques: Kinetic type
Camera: Static
14 · Problem · 0:46–0:47
A white diagram shows two horizontal waveform rows labeled CALLER and AGENT, with delegating pills and turn markers. A dark speech bubble holding a quote fades out as a Typical duplex model label appears at the top.
On-screen text: “Wait a second, let me think...”, “Typical duplex model”
Camera: Pull out
15 · Problem · 0:47–0:49
A pale diagram of two audio waveforms, labeled CALLER and AGENT, pans left under a gray badge and a dark headline. The headline “A brief pause.” fades in word by word while “delegating…” tags slide with the agent waveform.
On-screen text: “Typical duplex model”, “A brief pause.”, “and then I hear a brief pause when the”, “delegation is happening, and across many”
Techniques: Kinetic type
Camera: Static
16 · Problem · 0:49–0:52
On a white background, two grey audio waveform tracks labeled CALLER and AGENT scroll left under a 'Typical duplex model' badge, while a dark navy headline types out word by word across the top and then fades.
On-screen text: “Across many turns, that sounds robotic.”
Techniques: Kinetic type
Camera: Pull out
17 · Feature reveal · 0:52–1:03
Retell logo top center; a faint two-line waveform resolves into SPEECH and REASONING tracks under a navy serif headline that sharpens from blur; a playhead sweeps right, drawing a blue waveform and dotted pill as three numbered captions appear.
On-screen text: “SPEECH”, “REASONING”, “Interleaved reasoning.”, “Answers first”, “Reasons while it speaks”, “Finishes, grounded in the reasoning”
Techniques: Kinetic type
Camera: Static
18 · Feature reveal · 1:03–1:08
A man in a black shirt sits on a dark couch before floor-to-ceiling windows showing cypress trees, speaking to camera. A large white 1st fades in behind his head while three white lines stack in on the left under the Tell-1 logo.
On-screen text: “Tell-1”, “1st”, “productionized”, “interleaved”, “reasoning model.”
Techniques: Kinetic type
Camera: Static
19 · Feature reveal · 1:08–1:10
A white benchmark card lists four voice models with scores; the highlighted Tell-1 row slides from rank 4 up to rank 1. The card then fades out, leaving a small Reasoning OFF toggle at the top of a pale gradient background.
On-screen text: “vCX Voice AI Benchmark”, “which is the best model for Contact Center”, “With reasoning turned off, the model has”
Camera: Static
20 · Feature reveal · 1:10–1:25
A Reasoning toggle tops a pale gradient. A caller card links by a line to a one-shot agent card. The toggle flips on, the agent card becomes three option pills, two crossed out, the third blue with a check, then a line draws to a new agent card. Captions below.
On-screen text: “With reasoning turned off, the model has”, “to one-shot the answer.”, “It doesn't have any thinking beforehand”, “to ground the reasoning.”, “With thinking, the model can eliminate”
Camera: Static
21 · Feature reveal · 1:25–1:35
A man in a black sweater sits on a dark couch before windows with cypresses, talking and gesturing. White text builds word by word on the left with a book_appointment() badge, then a second line; captions run at the bottom and the clip cuts to a light chat screen.
On-screen text: “Execute the tool call at the right time.”, “book_appointment()”, “Solve complicated intention problems.”, “Like, for example, when you're calling,”, “you might have someone who has, like,”
Techniques: Kinetic type
Camera: Static
22 · Feature reveal · 1:35–1:44
The camera pulls back from a 'Reschedule appointment' card to reveal a grid of intent cards, each with a small SOP pill. A serif counter counts up to 1,000 as the caption 'intents, each with its own SOP' builds word by word; the Reschedule card then gains a blue outline.
On-screen text: “1,000”, “intents, each with its own SOP”
Techniques: UI zoom, Kinetic type
Camera: Pull out
23 · Feature reveal · 1:44–1:50
A man in a black t-shirt sits on a dark couch before a sunlit window, talking and gesturing. Two white chat bubbles labeled CALLER pop in at left and right with sample questions, while white subtitles run along the bottom.
On-screen text: “For example, a user could randomly ask,”, “CALLER What's the weather in Miami?”, “CALLER What's the banana bread recipe?”, “The model, without additional prompting,”
Camera: Static
24 · Feature reveal · 1:50–1:57
Two caller question cards and a blue 'No additional prompting' badge sit above three workflow cards sliding in from the edges. A blue arrow draws from the right to the Book appointment card, labeled 'Back to the task,' as the final step gains a border and a checkmark.
On-screen text: “No additional prompting”, “The model, without additional prompting,”, “Back to the task”, “is able to steer the user back to the”, “task at hand and”, “continue with the workflow.”
Camera: Static
25 · Feature reveal · 1:57–2:01
A young man in a black crewneck sits on a dark sofa, speaking off-camera in a dim room with blurred plants. On the left, a white bold headline builds word by word, then a blue nested-triangle icon draws in and a small caption appears at the bottom.
On-screen text: “There is an”, “impossible triangle.”, “people have been trying to aim for.”
Techniques: Kinetic type
Camera: Static
26 · Feature reveal · 2:01–2:07
A triangle diagram links three labels, Intelligence, Latency, Cost, around a glowing blue orb on a dark navy field with a faint serif 'Tél-1' watermark. The labels glow blue in turn, then checkmarks pop in. The shot opens with a triangular wipe over a blurred outdoor backdrop and cuts to a person.
On-screen text: “Within the models is”, “We've nailed all of them.”
Techniques: Masking
Camera: Static
27 · Feature reveal · 2:07–2:17
A man in a black shirt sits on a dark couch before floor-to-ceiling windows, gesturing as he speaks. A blue “Each customer” tag and a large white number that rolls to 1,000,000 with “calls per month.” appear over the shot.
On-screen text: “A regular voice agent, the cost might not”, “be the biggest factor, right?”, “But here, we're talking about like a”, “Each customer”, “1,000,000”, “million calls per month, kind”, “calls per month.”, “of volume for each customer,”, “and this is where you need to consider”
Techniques: Kinetic type
Camera: Static
28 · Feature reveal · 2:17–2:20
A young man in a black shirt sits on a dark couch, speaking to an off-camera interviewer. Soft daylight and blurred plants fill the background. White lowercase subtitles at the bottom swap from one line to the next as he talks. The frame stays locked off.
On-screen text: “and this is where you need to consider”, “the cost and how you scale here.”
Camera: Static
29 · Feature reveal · 2:20–2:27
A man in a black sweatshirt sits on a dark couch before tall windows with cypress trees, talking and gesturing. White overlay text builds line by line on the left, “Highly intelligent for contact center tasks.”, then on the right, “Latency is extremely low.”, with small bottom captions.
On-screen text: “the cost and how you scale here.”, “Highly intelligent”, “for contact center”, “tasks.”, “and latency is extremely low.”, “Latency is”, “extremely low.”, “It's the time-to-first token you see is”
Techniques: Kinetic type
Camera: Static
30 · Feature reveal · 2:27–2:35
A man in a black shirt sits on a dark sofa, speaking in profile against a blurred background. A millisecond timeline bar with a blue progress line runs along the bottom; white metric labels ~150ms and 200–250ms fade in on either side as the numbers count up.
On-screen text: “It's the time-to-first token you see is”, “time to first token”, “~150ms”, “around 150 milliseconds, and you can hear”, “the first audio at around like, you know,”, “first audio”, “200–250ms”, “200 to 250 milliseconds.”
Techniques: Kinetic type
Camera: Static
31 · Feature reveal · 2:35–2:38
A white card on the left reveals an isometric three-layer stack: the top slab fills from blue outline to solid blue with dots, above light blue and gray slabs. Below it, the serif label “Custom inference stack” fades in word by word beside a blue dot mark.
On-screen text: “First is our custom inference stack.”, “Custom inference stack”
Techniques: Kinetic type
Camera: Static
32 · Feature reveal · 2:38–2:44
A light gradient stage holds two cards. Left, a static white card shows an isometric blue layer stack titled 1 Custom inference stack. Right, a dark Retell panel draws an arc marked network hop with an X, then a glowing Tell-1 sphere pops in and a dotted line links it to the Voice agent pill as 2 Hosted in our production cluster types in.
On-screen text: “Custom inference stack”, “Second is we skip the network latency by”, “Hosted in our production cluster”, “hosting the model directly”, “in our production cluster.”
Techniques: Kinetic type
Camera: Static
33 · Other · 2:44–3:00
Static shot of a man in a black sweatshirt on a dark couch before tall windows, speaking to camera. A white headline builds word by word at left, then three blue pill labels pop in; a centered line follows as the frame dims and white dots scatter over his face.
On-screen text: “It's more like a professional agent”, “Like a professional agent.”, “helping you solve your tasks, or trying”, “Solve your task”, “to sell you something, or trying to like”, “Sell you something”, “book you an appointment, right?”, “Book an appointment”, “Those will be”, “different wording and style. The go-to model for any contact center use cases.”
Techniques: Kinetic type, Particles
Camera: Static
34 · Outro · 3:00–3:01
A faint silhouette of a seated person fades out to black, leaving a white ring of dots spinning in the center of the frame.
Camera: Static
35 · Outro · 3:01–3:03
A white Retell logo with a circular dot icon sits centered on a plain black background. The logo remains static throughout the short clip.
On-screen text: “Retell”
Camera: Static




