AI can now make videos of anyone. Pulse AI keeps the person consistent across all of them. Each person enrolls once with consent. Every video from any model is checked against that identity and blocked if it drifts.
RESEARCH
Working Paper Published (DOI)
OPEN METHOD
Evaluation Code on github
PRODUCT
Invite-only
STATUS
Operating
WORKING PRODUCT · REAL OUTPUT
This is real output from our pipeline: one enrolled identity, three approved scripts, three verified videos, delivered by email. No re-recording was required.
Watch with sound on.
STEP 1 · INITIAL CONVERSATION
"Hi Lucy, I noticed you're VP of Sales. Made this for you — worth 30 seconds?"
STEP 2 · FOLLOW-UP
"Following up on my note from last week. Still think this could change your reply rates, Lucy."
STEP 3 · AFTER THE CALL
"Good speaking earlier, Lucy. Two things I'd prioritize: the SDR script, then the handoff."
Run this on your pipeline
PRODUCT
Pulse AI turns one authorized sender identity into a connected sequence of personalized videos for your highest-value prospects.
Consented sender identity
You approve the source material once. Every message in the sequence is built under that authorization.
Sequence-aware personalization
Each touch references the prospect, their role, and where the conversation stands. Not a template with a name swapped in.
Reviewed before it ships
We review every output for sender credibility. Nothing enters your outbound without your approval.
THE IDENTITY LAYER
Generation models render a person anew with every output. We hold a single enrolled identity between the model and the message, and enforce four commitments on every output.
Consent
No signed release, no generation. This is enforced before the first frame.
Persistence
Every output is checked against one fixed enrolled identity, across the whole sequence. Today that check includes human review on every message.
Approval
Nothing reaches a recipient without the customer's approval. There are no automatic sends.
Revocation
A hard stop. Generation halts, hosted outputs are disabled and source material is deleted.
WHY PULSE AI
The cost of generation falls every quarter. The cost of a sender who appears even slightly wrong does not. Our role is to hold identity constant, so nothing inconsistent reaches a prospect.
Output
A single video
A connected sequence
Identity
Re-rendered in every video, unchecked
One enrolled identity, held across messages
Persistence
Inside one tool, unmeasured
Enrolled once; reuse across campaigns is next
Enrolled once; reuse across campaigns is next
Left to the user
Consent captured before generation
Measurement
Per-video quality
Every message checked against the enrolled identity
Revocation
Left to the user
Generation stops on request
WHY NOW ?
Cost
Generating a person is nearly free
Realistic video of a real person gets cheaper and faster every quarter. Volume is solved.
Law
Consent is becoming mandatory
Laws on biometrics and digital replicas now make consent mandatory.
Trust
Every recipient now asks
Deepfake fraud has made a new question routine: is this really them, and did they authorize it?
RESEARCH · SEQUENCE PERSISTENCE
A video can pass every quality check while the sender quietly changes between messages. Our published working paper introduces the Sequence Persistence Score (SPS), which evaluates whether one identity remains the same sender across an AI-generated sequence.
01
Enrolled identity
One enrolled identity, from consented source material. Every output is judged against it.
02
Sequence-level evaluation
Outputs are judged across the whole sequence (first touch, follow-up, after the call), not one video at a time
03
Open Method
Working paper published on Zenodo (CC BY 4.0)
HOW IT WORKS
01
Enroll once
The sender signs consent. Approved source material becomes their enrolled identity, once.
02
Generate and check
Each message is generated, then checked against the enrolled identity. Outputs that fail never reach the send queue.
03
Approve and Deliver
Checked messages are delivered by email with a hosted video link, only after your approval
CUSTOMERS
FOUNDERS · CEOS
Founder-led outbound
You still sell. Your presence reaches every prospect without hours lost to recording.
SALES LEADERS
Pre-demo follow-up
Personal reminders before discovery and demo calls. They reinforce why the meeting matters.
GTM TEAMS
Account-based outreach
Role-aware messages to multiple stakeholders. One sender identity. One campaign logic.
What we don’t take: one-off brand videos, consumer avatars, or any sender who has not authorized it.
VISION
THE IDENTITY LAYER FOR AI-NATIVE REPRESENTATION
Generated communication is becoming the default. The hard question is no longer whether AI can produce the message. It is who sent it, and whether they authorized it. Pulse AI is building the layer that answers that, across every interaction.
NOW · AVAILABLE
Identity-consistent sequences
Personalized video sequences for B2B outbound, built on one authorized sender.
NEXT · BUILDING TOWARD
Continuity across formats
One persistent identity across media, channels, and interactions.
LATER · DIRECTION
Authorized representation
Authorization, provenance, and delegated permissions for digital representations and AI agents.
Now is available as a pilot. Next and Later are where we are taking it, not shipped product.
PILOTS
A pilot runs on one live outbound workflow. Before anything is generated, we agree in writing on what success means.
One live workflow. One sequence.
Fixed in writing before we start.
30 days, tied to the agreed KPI.
A written readout of what earned replies and what did not.
FOUNDING TERMS
[Founding term not offered to later customers, e.g. pilot price held after conversion]
Start a Pilot
Send a short brief on your workflow, then book a call with me.
Every AI-generated message will need a persistent, verifiable identity behind it. We are building that infrastructure on measurement, not assumption.
Ananya Das
Founder & CEO, Pulse AI
CONSENT & DATA
An identity company is judged by how it handles a face. These rules hold on every pilot. No exceptions.
Consent before generation
We generate nothing until the sender signs consent.
Scoped use
We use source material only for the sequences the sender approves.
Defined retention
We delete source material 30 days after the pilot ends, or the day you ask.
Revocable at any time
Withdraw authorization and generation stops.
DESIGN PARTNERS
A 30-day deployment on your live outbound, at fixed terms. Nothing is delivered without your approval.
FAQ