Seed · 2026

The production OS
for book publishers.

One manuscript in → every production output out — approved and connected in one title record.

Live with Yedioth Books · 83/87 AI findings accepted by a professional editor · in Israel as Seffer

Selected for Google for Startups

Seed · 2026

The production OS for book publishers.

One manuscript in → every production output out — approved and connected in one title record.

Live with Yedioth Books · 83/87 AI findings accepted by a professional editor · in Israel as Seffer

Selected for

The problem

02

Book production was never turned into software.

The same chain — for every title

The consequence

A chain of 5–8 specialists in separate tools.

2–6 months to market

Coordinated over email, spreadsheets and loose files.

20–50 handoffs & file versions

Production knowledge stays with individuals.

Nothing compounds to the next title

Publishing has systems of record — not a system of production. So every new title is rebuilt from scratch.

Book production was never turned into software.

2–6 moto market — every title rebuilt by 5–8 specialists across 20–50 handoffs.

Publishing has systems of record — not a system of production.

Impria · ConfidentialSlow and expensive — it’s still manual

Why now

03

AI crossed the book-length threshold. Approval workflows made it publishable.

Technology · new

Book-length AI became viable

Modern models process a full manuscript as a connected whole — at negligible cost: $174 total AI spend across 444 editorial runs to date.

Deployability · new

Approval made it professional

Raw model output was never professionally accountable. Structured approval — findings the publisher reviews, accepts or rejects — turns AI work into production an editor signs off on.

Demand · customer discovery

Capacity is the constraint

Six of six publishers we approached cited production capacity as the #1 constraint: output must grow without growing headcount.

The result: sequential production becomes one parallel software run — a first reviewable package in hours, not weeks, with the publisher approving every output.

AI crossed the book-length threshold. Approval made it publishable.

1
Book-length AI became viable
A full manuscript as a connected whole — $174 total AI spend across 444 runs
2
Approval made it professional
The publisher reviews, accepts or rejects every finding
3
Capacity is the constraint
Cited #1 by six of six publishers we approached

Sequential production becomes one parallel run — a reviewable package in hours, not weeks.

Impria · ConfidentialWhy this is possible now

Traction · anchor customer

04

One of Israel’s largest publishers

Yedioth expanded from one workflow to four — in six weeks.

Ma’ayan Zigdon

Editor-in-Chief
Yedioth Books

“

“With a deep understanding of the publishing process, Seffer addressed every pain point I had encountered — from its remarkable manuscript analysis and streamlined editing process to practical guidance on marketing strategies and costs.

Using the platform felt like playing with magic. I genuinely felt I was witnessing the future of publishing.”

Verified today paid & live

4

workflows delivered
editorial · covers · files · channel prep

10

titles processed
through the platform in ~6 weeks

~$3.3K

per month, paid
paying since June 2026

3

seats in use
Yedioth’s team, in-product

Next milestone Annual platform agreement in negotiation

Yedioth expanded from one workflow to four.

Ma’ayan Zigdon

Editor-in-Chief, Yedioth Books

“With a deep understanding of the publishing process, Seffer addressed every pain point I had encountered — from its remarkable manuscript analysis and streamlined editing process to practical guidance on marketing strategies and costs.

Using the platform felt like playing with magic. I genuinely felt I was witnessing the future of publishing.”

4workflows delivered — editorial · covers · files · channel prep · 10 titles in ~6 weeks · paying since June 2026 · team in-product.

Next milestone: annual platform agreement — in negotiation.

Measured proof · real titles

05

A 100K-word title reached a reviewable production package in 2 hours — not 3 weeks.

2 hrs

to first reviewable analysis · vs ~3 weeks manually — publisher-reported baseline

83/87

findings accepted by Yedioth’s editor

17

review tasks — 87 findings consolidated for the editor

10 min

to production-ready EPUB + print-ready PDF

Editorial findings located inside the manuscript

Findings inside the manuscript — accept / reject. Live product · Hebrew UI.

Finished EPUB export from the same title record

Ten minutes later — the sellable EPUB, from the same title record.

100K words → reviewable package in 2 hours.

2 hrsvs ~3 weeks manually (publisher-reported) · 83 of 87 findings accepted by Yedioth’s editor → 17 review tasks · sellable EPUB + print-ready PDF in 10 min.

Impria — exploded book, manuscript to market

Product · the production engine

06

One manuscript in.
Every production output connected.

The engine does the work.

The publisher approves every output.

1

Editorial intelligence

Findings → reviewable tasks. The publisher approves every one.

2

Production assets

Cover, metadata, EPUB & print-grade PDF — publisher-approved.

3

Channel-ready files & title data

Prepared to each channel’s spec. Every output stays connected in the title record.

Measured results on the previous slide · full per-function economics in the appendix.

One manuscript in. Every output connected.

One manuscript becomes every connected production output

1
Editorial intelligence
Findings → review tasks; the publisher approves
2
Production assets
Cover, metadata, EPUB & print-grade PDF
3
Channel-ready files & title data
Prepared to spec; every output stays connected in the title record

Business model

07

High-ACV publisher software that expands with title volume.

Small publisher proposed pricing

Base platform$12K

+ 25 titles × $1,000$25K

Annual contract value$37K

Mid-market publisher proposed pricing

Base platform$24K

+ 85 titles × $750$63.75K

Annual contract value~$88K

Large publisher proposed pricing

Base platform$60K

+ 300 titles × $500$150K

Annual contract value$210K

Today live

~$3.3K / mo actual, paying since June 2026 — a limited deployment (10 titles · 4 workflows) that annualizes to ~$40K. Full-catalogue annual deployment prices at ~$88K — tiers are set by annual title volume, and Israel’s largest publishers produce at US mid-market scale (~50–199 titles/yr). Annual agreement in negotiation.

The expansion story

More titles → more modules → more seats → higher ACV.

Future expansion: freelancer marketplace (~$1.7K avg gig · 5% take) — economics in the appendix.

Full function-by-function economics — nine functions, per-function savings — in the appendix.

Software that expands with title volume.

Base platform + per-title usage · proposed pricing.

Small publisher$37K

Mid-market publisher~$88K

Large publisher$210K

Marketplace · later · appendix~$1.7K

Market

08

Thousands of publishers still buy production as labor — not software.

The goal: the infrastructure through which every book is produced, approved and delivered.

The venture math · per-title capture

~$15–30K→$0.7–1.5K

Labor spent per title today (modeled) → Impria software + usage revenue per title at proposed pricing. Revenue scales with every title produced — ~5% of the spend, before marketplace take-rate.

~$1–2B / yr

The US pool that spend sums to — ~70K titles/yr across 1,000 qualified accounts (of ~2,300 Census firms). UK & Europe (~4,000+ accounts, est.) extend it.

How Impria captures the pool

Today · software + per-title usage

1,000 qualified US accounts × $37–210K ACV = $69.5M software beachhead (modeled) — the entry point, not the ceiling · $200M+ with UK & Europe (directional)

The position · production infrastructure

Every title is produced through the platform — revenue scales with titles produced, not seats sold.

Long-term · take-rate on the pool itself

Freelance & services production spend routed through the platform (marketplace · 5% take · appendix) — GMV economics on the $1–2B, not license fees.

Sources: AAP StatShot 2024 · 2023 US Census CBP (NAICS 51113) · UK Publishers Association · bottom-up, directional; tier counts are management estimates.

A labor market becoming a software market.

$15–30K labor/title → $0.7–1.5K to Impria · a ~$1–2B/yr US pool (modeled)

1
Today · software + usage
1,000 US accounts × $37–210K → $69.5M beachhead — the entry point · $200M+ with UK & EU
2
The position · infrastructure
revenue scales with titles produced, not seats sold
3
Long-term · take-rate
GMV economics on the $1–2B pool itself (marketplace · 5% take)

Go-to-market

09

Land on one title. Expand across the publisher.

The motion

1

Land on 1–3 live titles.

2

Prove measured time & cost savings.

3

Convert to an annual platform + per-title usage.

4

Expand across the whole list.

How a pilot runs

Buyer: Editor-in-Chief / Head of Production.
A ~4-week paid pilot on live titles.
Converts on measured time & cost savings.

Pipeline today: 1 paying deployment · 3 active pilots (paid & unpaid mix) · 3 further qualified conversations

→ 18-month objective: 15–20 paying publishers

One platform, two brands — Impria internationally · Seffer in Israel.

Founder-led sales into mid-sized independents · first dedicated US sales hire funded by this round · flagship references from live deployments.

Land on one title. Expand across the publisher.

1
Land
on 1–3 live titles — a real bottleneck
2
Prove
measured time & cost saved
3
Convert
to an annual platform + usage
4
Expand
across the whole list

1 paying · 3 active pilots (paid & unpaid mix) · 3 qualified conversations → 15–20 paying publishers. One platform, two brands: Impria internationally · Seffer in Israel.

Competition · moat

10

Publishing has systems of record. It still lacks a system of production.

Systems of record exist. A system of production doesn’t.

Incumbents manage metadata around the manuscript. Impria works inside it — and produces the downstream assets.

Team

11

We mapped publishing end-to-end — then built it.

Full-time founding teamProduct liveAnchor publisher paying

Dvir Tzur

CEO · PhD Literature, HUJI

20 yrs literary editor · published author.

Yonatan Ovadia

COO & Legal · Lawyer, HUJI

15 yrs publishing rights & ops.

Elinadav Heymann

CTO & CPO · MA, HUJI

20 yrs programming & IT · built Impria’s production architecture, long-document AI orchestration and print-grade publishing engine — end-to-end.

Editorial craft · publishing operations · systems engineering — the three disciplines to turn book production into software.

We mapped publishing end-to-end — then built it.

Dvir TzurCEO · PhD Literature, HUJI — 20 yrs literary editor, published author

Yonatan OvadiaCOO & Legal · Lawyer, HUJI — 15 yrs publishing rights & ops

Elinadav HeymannCTO & CPO · MA, HUJI — 20 yrs programming & IT; built Impria’s production architecture & AI orchestration end-to-end

Financial plan · 3-year model

12

The floor scenario: on $2M alone, this company cannot die.

US operating model

Y1 · 2027

Y2 · 2028

Y3 · 2029

Revenue

$363K

$1.12M

$2.64M

Exit ARR

$634K

$1.56M

$3.6M

Paying publishers (ending)

~12

~25

~49

Gross margin

88%

EBITDA

−$693K

−$587K

+$138K

Ending cash

$1.27M

$603K

$572K

EBITDA-positive ~month 30Cash never below ~$0.4M36-month US model · three tiers

This is the no-further-capital case — EBITDA-positive at ~month 30 with the seed alone. A Series A doesn’t rescue this plan; it accelerates the same motion: more sales capacity into the same 1,000-account beachhead and the per-title capture on the $1–2B pool.

Operating plan — expected-value scenario, not guidance. Full monthly model available in diligence.

The floor scenario: on $2M alone, this company cannot die.

$3.6MY3 exit ARR — $2.64M Y3 revenue · ~49 paying publishers · 88% gross margin.

M30EBITDA-positive ~month 30 — cash never below ~$0.4M across the 36-month model.

Operating plan — expected-value scenario, not guidance.

The ask

13

Raising · Seed

$2.0M

on a SAFE — to repeatable international sales: 15–20 paying publishers · ~$1M ARR

Use of funds · through EBITDA-positive

40%Go-to-market

35%Product & engineering

10%Implementation & customer success

15%Ops & admin

36-month model runs EBITDA-positive at ~month 30 — financial plan, p. 12.

What this round must prove · 18 months

Commercial repeatability

15–20 paying publishers → $60–100K MRR (e.g. 12 small + 6 mid-market ≈ $81K at proposed pricing)

Product economics

Gross margin ≈ 88% (modeled) · verified time & cost savings across ≥20 titles

Customer expansion

≥50% of contracted publishers running 3+ production workflows

International

3 paid US/UK pilots under Impria · ≥1 international annual contract

Yedioth annual agreement in negotiation. One engine, two markets.

Raising · Seed

$2.0M

on a SAFE — to repeatable international sales: 15–20 paying publishers · ~$1M ARR.

Go-to-market40%

Product & engineering35%

Implementation & success10%

Ops & admin15%

This round proves 15–20 paying publishers · $60–100K MRR · ~88% margin. One engine, two markets.

Appendix · function by function

14

Nine functions publishers buy as labor. Impria runs them as software.

Function

The old way

With Impria

Addressable labor spend · modeled

What it means for Impria

Production coordination live

A coordinator chases specialists over email & spreadsheets — 20–50 handoffs per title.

Every function, status and file connected in one title record.

~30 hrs → ~3 hrs~90% · $1–1.5K

Connective tissue rivals can’t retrofit — switching cost deepens.

Editorial passes verified

Four separate hires — developmental, line, copyedit, proofread — each a pass, weeks apart.

One run covers the developmental, style & proofing layers in ~2 hrs — 95% editor-accepted.

~3 wks → ~2 hrs~90% · $8–16K

Repeatable half of four editorial roles → per-title revenue + switching cost.

Cover & design live

Brief a designer, then wait days for each revision round.

Cover concepts generated inside the title record; the publisher approves.

~1–2 wks → hrs~85% · $1–2K

A purchased service becomes a platform feature; custom work to marketplace.

Typesetting & files verified

A typesetter rebuilds the book for every format — weeks, new errors.

Print-ready PDF & EPUB in ~10 min, from the same record.

~1–2 wks → ~10 min~95% · $0.4–1.5K

Near-zero marginal cost per format — the 85%+ gross-margin target.

Metadata & catalog live

Retyped by hand into every channel’s forms and spreadsheets.

Written once in the title record, reused across every output.

~4 hrs → mins~90% · $150–300

Impria holds the canonical title record trackers never see.

Legal & contracts live

Author & rights contracts drafted from scratch, emailed back & forth with counsel.

Contracts drafted from the publisher’s choices, ready for in-house counsel to approve.

~days → same-day~85% · $0.8–2K

Impria produces the contract; counsel signs — records only store it.

PR & press live

A publicist writes copy, designs materials, works the media list and chases coverage.

All PR graphics generated in seconds; copy & press materials prepared for the media list.

~2–4 wks → days~90% · $3–8K

A whole agency line-item absorbed — more spend per title on Impria.

Distribution live

Separate accounts, file specs and uploads for each platform.

One package prepared for 20+ platforms and 13 Amazon markets.

~1–2 wks → hrs~90% · $0.3–0.5K

Impria sits between publisher and channel — usage revenue per title.

Specialist sourcing planned

A private Rolodex per publisher; days to find, brief and pay each freelancer.

Work that still needs a human is routed to vetted freelancers in the workflow.

~days → minsadmin cut · proj.

Talent gathers where paid work flows — network becomes the moat.

~90% modeled of the ~$15–30K addressable labor spend per title → ~$1.5–3K · months → days to market

Verified = measured on live publisher titles · live = shipped, not yet economically measured · planned = roadmap. Savings modeled from standard US freelance / agency rates; figures vary by title length and segment. AI grounding: 444 editorial runs · 5,167 line suggestions · ~$174 total AI spend to date. Marketplace (modeled): ~900 engagements ≈ 18 publishers × ~25 titles × 2 per title · ~$1.5M GMV · 5% take. Rights & IP: AI cover output is concept material for human designer refinement — final cover art is human-designed; editorial output is suggestions a human editor accepts or rejects — the accepted edit is the publisher’s. Model dependency: AI compute is ~4% of revenue (modeled) — even a 4× provider price increase leaves gross margin ~76%.

Labor in. Software out.

Nine functions publishers buy as labor — run as software. ~90% modeled of the ~$15–30K addressable labor spend per title → ~$1.5–3K, months → days.

1
Production coordination
20–50 email handoffs → one connected title record · ~30 hrs → ~3 hrs, ~90%
2
Editorial passes
Dev · line · copyedit · proofread → one ~2-hr run · ~3 wks → 2 hrs, ~90% ($8–16K)
3
Cover & design
Days per revision → concepts in the title record · ~1–2 wks → hrs, ~85%
4
Typesetting & files
Weeks per format → EPUB + print PDF in ~10 min · ~95%
5
Metadata & catalog
Retyped per channel → written once, reused · ~4 hrs → mins, ~90%
6
Legal & contracts
Drafted from scratch → generated from choices, counsel approves · ~85%
7
PR & press
All PR graphics in seconds; copy & press materials prepared · ~2–4 wks → days, ~90% ($3–8K)
8
Distribution
Per-platform uploads → one package prepared for 20+ platforms · ~90%
9
Specialist sourcing
Private Rolodex → vetted freelancers in the workflow · planned

Each function becomes software revenue — and the freelancer network becomes the moat. Editorial & file outputs measured, other savings modeled from freelance / agency rates · direct distribution & specialist routing planned.

Impria · ConfidentialLabor in, software out — every function

Appendix · files & distribution

15

One title record becomes every sellable format.

The unified title record — live dashboard

The unified title record — tasks, team & activity for one live title.

Full-Hebrew EPUB + cover → retailers & libraries.

Print-ready PDF typesetting in minutes, not weeks.

Market-ready across major channels

20+ platforms · 13 Amazon markets · ebook + paperback. All from one title record. Files & metadata prepared to each channel’s spec — uploads founder-assisted today; direct API distribution on the roadmap.

One title record → every sellable format.

The unified title record — tasks, team & activity for one live title.

Full-Hebrew EPUB + print-ready PDF, generated in minutes.

20+ platforms · 13 Amazon markets · ebook + paperback · files prepared to spec; uploads founder-assisted today.

Appendix · US market by tier

16

The $69.5M US beachhead, tier by tier.

Tier

Annual titles

US accounts

Proposed pricing

Base ACV

US TAM

Small

10–49

600

$12K platform + $1,000/title × ~25

$37K

$22.2M

Medium

50–199

300

$24K platform + $750/title × ~85

$87.75K

$26.3M

Large

200+

100

$60K platform + $500/title × ~300

$210K

$21M

US total

1,000

$69.5M

Annual ACV expansion 5%Churn 5% / 3% / 2% by tierUK & Europe: ~4,000+ further accounts (est.)

Qualified accounts are management estimates from 2,347 US book-publisher establishments (2023 US Census CBP, NAICS 51113), excluding micro-publishers and duplicate buying entities. ACV = platform fee + per-title fee × average titles per account. Full monthly model available in diligence.

The $69.5M US beachhead, tier by tier.

1
Small · 600 accounts
$12K + $1,000/title × ~25 → $37K ACV · $22.2M
2
Medium · 300 accounts
$24K + $750/title × ~85 → $87.75K ACV · $26.3M
3
Large · 100 accounts
$60K + $500/title × ~300 → $210K ACV · $21M

US total: 1,000 accounts · $69.5M. From 2,347 Census establishments; UK & Europe on top.

The production OSfor book publishers.

The production OS for book publishers.

Book production was never turned into software.

Book production was never turned into software.

AI crossed the book-length threshold. Approval workflows made it publishable.

Book-length AI became viable

Approval made it professional

Capacity is the constraint

AI crossed the book-length threshold. Approval made it publishable.

Book-length AI became viable

Approval made it professional

Capacity is the constraint

Yedioth expanded from one workflow to four — in six weeks.

Yedioth expanded from one workflow to four.

A 100K-word title reached a reviewable production package in 2 hours — not 3 weeks.

100K words → reviewable package in 2 hours.

One manuscript in.Every production output connected.

One manuscript in. Every output connected.

Editorial intelligence

Production assets

Channel-ready files & title data

High-ACV publisher software that expands with title volume.

Software that expands with title volume.

Thousands of publishers still buy production as labor — not software.

A labor market becoming a software market.

Today · software + usage

The position · infrastructure

Long-term · take-rate

Land on one title. Expand across the publisher.

Land on one title. Expand across the publisher.

Land

Prove

Convert

Expand

Publishing has systems of record. It still lacks a system of production.

Systems of record exist. A system of production doesn’t.

We mapped publishing end-to-end — then built it.

Dvir Tzur

Yonatan Ovadia

Elinadav Heymann

We mapped publishing end-to-end — then built it.

The floor scenario: on $2M alone, this company cannot die.

The floor scenario: on $2M alone, this company cannot die.

Nine functions publishers buy as labor. Impria runs them as software.

Labor in. Software out.

Production coordination

Editorial passes

Cover & design

Typesetting & files

Metadata & catalog

Legal & contracts

PR & press

Distribution

Specialist sourcing

One title record becomes every sellable format.

One title record → every sellable format.

The $69.5M US beachhead, tier by tier.

The $69.5M US beachhead, tier by tier.

Small · 600 accounts

Medium · 300 accounts

Large · 100 accounts

The production OS
for book publishers.

One manuscript in.
Every production output connected.