Status · Proposal · ready to review
Phones · 19 frames across 7 chapters
Companion · this document is self-contained
Back to overview
Google Flow Beta has three structural opportunities, in increasing order of leverage. One — the packaging routes every upgrade into Google One, forcing the image/video persona to subsidise storage and Gemini-in-Gmail features they don't want. Two — the playground is thin: prompt + ratio + count + duration + model picker is the entire creator surface, while every serious peer ships 3–6 more capability classes. Three — the most Google-native workflow on the table, turning a NotebookLM into a full video, isn't built yet. This proposal addresses all three: standalone pricing, a scalable playground that closes the feature gap, and a flagship workflow agent built on NotebookLM + Gemini + Veo + Imagen, all Google-owned.

The forecast

At 90 days post-launch: paid conversion +38 % (1.7 % → 2.3 %), blended ARPU +22 % (₹1,460 → ₹1,780), 30-day retention +10 ppts (64 % → 74 %), Library DAUs +60 % among users with 30+ assets, and a new acquisition channel: 8 % of new accounts arriving via NotebookLM cross-links — users (teachers, founders, authors) who would never have hit the prompt-driven funnel.
01
The problem · Beta today
02
The landing · 2 frames
03
The playground · 5 frames
04
From a Notebook · 3 frames
05
Google handoffs · 5 frames
06
Trust & safety · 2 frames
07
Pricing redo · 2 frames
08
Impact & rollout
09
Scalability · 6 axes
10
Maturity matrix vs peers
Chapter 01

The problem: Flow Beta is bundled, bare, and behind.

Today's Flow Beta has 5 generation modes, a flat Projects list, no search, no post-gen edit, no Character library, no multi-shot, no multi-ref image edit, no voice prompt, no native publish. Upgrade routes the user into Google One AI Plus / Pro / Ultra — plans where 70 % of the headline price funds storage, NotebookLM and Gemini-in-Gmail that the generation persona doesn't value.

What you actually buy when you upgrade Flow Beta today

PlanPrice (India)What the page leads withWhat the Flow user actually needs
AI Plus₹399 / mo200 GB cloud storage · Nano Banana in Search · Flow MusicJust more Flow credits and Veo Fast
AI Pro₹1,950 / mo5 TB storage · NotebookLM Plus · Gemini in Gmail · Veo QualityThe advanced creator tools — which don't exist yet
AI Ultra₹6,500 / mo20 TB storage · Deep Think · Project Mariner · Veo QualityPooled credits + brand kits — also absent

The competitive band creators actually compare against

ProductEntry paidPro tierDifferentiator
Kling 3.0$6.99$25.99Motion Control · Multi-Shot (6 stitched, 4K 60fps) · Avatar 2.0
Runway Gen-4$12$76Aleph post-gen prompt-edit · Act-Two performance · Motion Brush 3.0
Pika 2.5$8$28Pikaframes (keyframes) · Pikaformance (lip-sync) · Scene Ingredients
Sora 2$20$200Cameos (face-verified) · Storyboard · Remix (timestamped edit)
Grok Imagine$10$30Extend-from-Frame · Modify (NL video edit) · 14 aspect ratios
Krea$9$29Realtime Canvas (<50 ms regen) · 64+ models · LoRA training · 22K upscale
Higgsfield$9$29Cinema Studio · @-tagged characters · multi-model routing per shot
Flow Beta (today)$19.99 bundled$249 bundledNone — and the upgrade page sells you storage
Chapter 02 · 2 frames

The landing: first three seconds decide whether the user generates or bounces.

Beta drops users on a 5-mode strip and a timestamped Projects list. v1.0 grows the mode set to 8, surfaces From a Notebook as a hero card, and turns Projects into a real Library with search, tags and smart collections.
2A · Home — the mode set + Notebook hero + Libraryexpanded modes
12:36
📶📡
86
Google Flow
BETA
D
Generate
image
Generate
video
Animate
photo
Edit
photo
Edit
video
Storyboard
scenes
Lip-sync
& voice
Character
library
NEW
From a Notebook via NotebookLM
Turn a NotebookLM into a full storyboard — Gemini extracts the flow, you confirm the plan.
Library
Search prompts, characters, tags
All Videos Images Storyboards Aarav 9:16
May 24, 12:36 pm VIDEO · 8s · 9:16
Jagganath IMAGE · 1:1
Same dark canvas, same Beta header. Three layered changes: (a) mode carousel grows 5→8 with Storyboard, Lip-sync, Character library; (b) From a Notebook gets a featured row of its own — Flow's flagship workflow agent, powered by NotebookLM ingestion + Gemini storyboard extraction; (c) "Projects" becomes Library with search, filter chips, and metadata chips on each card.

How peers land users

  • Sora 2 app — vertical algorithmic feed of others' clips with Remix on every card. Engagement collapsed 45 % MoM by Jan 2026.
  • Kling app — tile grid (Image-to-Video, Motion Brush, Multi-elements, Lip Sync, Effects) + community feed + templates.
  • Runway — tools grid (Generate Video, Edit with Aleph, Act-Two, Voice, Lip Sync) + asset library with folders and search.
  • Higgsfield Cinema Studio — project board with @-tagged characters, locations and props as first-class assets.
  • Krea — Realtime Canvas as the primary playground (sub-50 ms regen as you type or sketch).
What Flow v1.0 changes
  • Mode grid grows 5 → 8: Storyboard scenes, Lip-sync & voice, Character library
  • "From a Notebook" gets its own hero row — Flow's flagship workflow agent
  • Projects → Library with search bar, filter chips (type / ratio / character / recency) and metadata badges on every card
What we deliberately don't add

No public social feed, no community gallery, no leaderboards. Sora 2 proved the feed model is brittle for AI video; Google has YouTube Shorts for the social layer (chapter 5).

Library at scale

Beta's flat Projects list scales for ~10 items. At 50+ saved generations users need search, tags and smart collections — every peer has them.

What's on this screen
  • Smart collections auto-grouped — Aarav (12), Monsoon (8), Brand kit (23). No folder discipline required from the user.
  • Search over prompts and tags — Grok Imagine's tag-filter model, adapted
  • Filter chips — type, aspect ratio, model, recency, sort
  • Card thumbnails carry metadata badges (VIDEO · 8s · 9:16)
  • Generation history per project, with branch & revert
KPI

Library DAU among users with ≥ 30 assets: +60 % at day 90. The product only becomes usable past 30 generations when search and tags exist.

2B · My Library — search, smart collections, tagstags + smart collections
12:42
📶📡
85

My library

Smart collections
Aarav
12
Monsoon
8
Brand kit
23
monsoon 3 results
Videos 9:16 Veo 3.1 Last 7 d Sort ▾
8s · v3
Mumbai monsoon
8s · Aarav
Cafe scene cut
6s
Pull-back · v2
Beta's flat Projects list doesn't scale past 30 items. Library replaces it with smart collections (auto-grouped by character, theme, brand kit), search over prompts and tags, and filter chips (type, ratio, model, recency, sort). Direct response to Grok Imagine's tag filter and Runway/Krea's folder model — Flow gets it without forcing folder discipline.
Chapter 03 · 5 frames

The playground: every peer has 3–6 more capability classes than Flow.

Beta exposes prompt + ratio + count + duration + model. v1.0 adds motion brush, keyframes, camera path, references, character library, seed lock, negative prompt, multi-ref image edit, multi-shot stitching, and a full post-gen Aleph-class toolbar. Peers (Runway, Kling, Pika, Sora, Grok, Krea, Higgsfield) all expose most of these.
3A · Generate video — every advanced control made first-classexpanded controls
12:37
📶📡
86

May 24 · 12:37 pm

No assets yet — create one or upload from your gallery.
Cinematic dolly-in on a vintage red car parked outside a Mumbai cafe, monsoon rain, neon reflections…
Video · x1
Generating will use 20 AI credits

16:9

9:16
x1
x2
x3
x4
4s
6s
8s
10s
🍃 Veo 3.1 — Fast
Seed 4172 locked
Negative…
Motion brush
Keyframes
Camera
Refs/Char
Extend
Style
Lip-sync
History
The Beta only exposes ratio, count, duration, model. We add 8 first-class creator tools — Motion brush, Keyframes, Camera path, Refs/Character, Extend, Style preset, Lip-sync, History — plus seed lock (re-roll identical conditions, +1, +2) and negative prompt. Peers (Runway Motion Brush, Pikaframes, Kling Motion Control + start/end frames, Sora Cameos, Grok Extend-from-Frame) all expose these.

Eight advanced tools, plus seed lock + negative prompt

The Beta creator panel has one model picker. v1.0 adds a second row with seed lock (re-roll identical conditions, +1, +2) and negative prompt, then eight first-class creator tools below:

  • Motion brush — region-targeted motion (Kling, Runway)
  • Keyframes — start/end frames (Pikaframes, Kling)
  • Camera — preset paths (dolly, orbit, pull-back)
  • Refs/Char — drop in up to 14 reference images
  • Extend — Grok Imagine's Extend-from-Frame pattern
  • Style — brand kits, look-dev presets
  • Lip-sync — Pikaformance / Avatar 2.0 / Act-Two parity
  • History — per-project revision tree, branch & revert
Tier

Motion brush, Keyframes, Style, Refs, Seed, Negative ship at Creator. Character library, Camera, Lip-sync, Storyboard, Extend ship at Pro.

Image playground — the modality Beta doesn't have at all

Flow Beta has "Edit photo" mode but no real image-editing surface. v1.0 ships the missing one: inpaint with brush, outpaint, remix, plus up to 14 reference images (matching Gemini 3 Pro Image's published limit; GPT Image 2 supports 16; Grok Imagine 3).

Engines

Powered by Nano Banana Pro (Google's own) and Imagen 4 — both best-in-class for text-in-image as of May 2026. Not a model aggregator like Krea.

Why this exists

Image is the highest-volume modality for creators — brand assets, social cards, posters, product shots. Beta forces them to use other tools and re-upload. v1.0 keeps them inside Flow.

3B · Image playground — multi-ref, inpaint, outpaint, remixmulti-ref compose
12:43
📶📡
85

Edit image

🍌 Nano Banana Pro
brush · 32px
+
Replace dress with the red one in ref 1, keep face from ref 2…
3 refs
Brush
Lasso
Relight
Upscale
4K
The image playground Flow Beta doesn't have at all. Inpaint with a brush, outpaint, remix, plus up to 14 reference images (matches Gemini 3 Pro Image's published limit). Closes the gap with Krea's canvas, Runway's Aleph for images, and Grok Imagine's 3-image compose — but powered by Nano Banana Pro and Imagen 4, both already best-in-class for text-in-image.
3C · Asset picker — Photos + Characters + Styles+ Characters & Styles
12:37
📶📡
86
Google Flow BETA
Videos Collections Characters Styles
Flow will only have access to the photos that you select
Dismiss
00:20
00:26
00:30
00:30
00:30
00:22
Saved characters
NEW
Aarav · v3
Riya · v1
+ Train new
Bahubali brand
The existing Google Photos sheet ("Videos / Collections" + consent) stays — we add Characters and Styles tabs, plus a "Saved characters" rail. Lets users reuse trained likenesses across projects (table stakes vs Sora Cameos and Higgsfield).
3D · Storyboard scenes — multi-shot stitchingnew mode
12:40
📶📡
85

Mumbai monsoon · story

3 shots · 24s
8s
Shot 01
Dolly-in on red vintage car, neon café reflections, monsoon rain
transition: cross-dissolve
8s
Shot 02 · Aarav v3
Aarav steps out, lights cigarette, deep amber lighting
transition: match-cut
Shot 03 · generating · 14s
Slow pull-back · rain stops · sunlight breaks through
Character consistency: Aarav · v3 edit
Stitches up to 6 shots into a continuous narrative with shared characters and transitions — closing the gap to Kling 3.0 Multi-Shot and Sora 2 Cameos. Each shot can use a different model; the character (here "Aarav v3") propagates so the face stays consistent.

Asset picker & Storyboard — character continuity made first-class

The asset picker keeps the existing Google Photos sheet (read-only, consent-gated — unchanged from Beta) and adds two tabs: Characters (trained likenesses, chapter 6) and Styles (saved brand looks). Storyboard mode then strings up to 6 shots together with explicit character-consistency cues — same pattern as Kling Multi-Shot, Higgsfield Cinema Studio scenes and Sora 2 Storyboard.

3E · Result & post-gen edit — Aleph-class toolbarAleph-class tools
12:38
📶📡
85

Mumbai monsoon · v3

0:08
SynthID · C2PA
Use as next shot YouTube Shorts Download Share Drive · opt-in
Edit this clip Aleph-class · Pro
Edit with
prompt
Remove
object
Relight
Add
sound
Upscale
4K
Extend
8s →
Voiceover
Caption
+ translate
Where Flow's native handoffs shine. Top chip row leads with YouTube Shorts publish (kept — Google owns YouTube, 9:16 ready, one-tap upload with provenance pre-attached) and "Use as next shot" (storyboarding). Google Drive is muted and opt-in. Provenance badge in the corner — SynthID + C2PA — is shipped by default on every Flow output. Aleph-class edit toolbar (Pro tier) below; Caption + translate added for Gemini stack integration.

Post-generation toolbar — Flow's biggest miss

Today after a generation completes, Flow Beta shows Play / Download / Share. That's it. Every peer ships an Aleph-class edit step. v1.0 ships eight:

  • Edit with prompt — Runway Aleph parity
  • Remove object — Pikaswaps parity
  • Relight — Aleph relight
  • Add sound — SFX, ambience, music
  • Upscale 4K — Kling/Runway/Krea parity
  • Extend 8s → — Grok Extend-from-Frame
  • Voiceover — Gemini TTS voices
  • Caption + translate — Gemini stack, native to Google
The top action row

Leads with Use as next shot (storyboard flow), then the prominent red YouTube Shorts chip (chapter 5), then Download & Share, with Google Drive muted to opt-in. The corner badge — SynthID · C2PA — ships on every output by default (chapter 6).

Chapter 04 · 3 frames · flagship

From a Notebook: the category nobody else can ship.

Every peer starts from a prompt or a reference asset. Flow + NotebookLM starts from a grounded knowledge source. The user already has the material — a kids' storybook, a pitch deck, a brand book, a textbook chapter, a screenplay. NotebookLM has been the place where users assemble that material since 2024. The missing step is turning it into video — and only Google owns both halves of the stack.
4A · From a Notebook — pick existing or attach sourcesNotebookLM ingestion
12:48
📶📡
83

From a Notebook

Your notebook → a full Flow

Pick an existing NotebookLM, or attach sources now. Gemini reads it, proposes a storyboard, you confirm — Flow renders the whole thing in one go.

Your notebooks
Tara & the rain — a kids' story
4 sources · 8 sections · updated yesterday
Attive Q2 pitch + brand book
12 sources · 14 sections · 3 d ago
Class 7 Geography — Monsoons
3 sources · 6 sections · 1 wk ago
Or attach sources now
PDF · DOC · image · URL · audio
Creates an inline notebook · stays in Flow
Two ways in. Either pick an existing NotebookLM (Flow lists the user's notebooks via NotebookLM integration — sources, sections, last-updated all visible) or attach sources inline (PDF / doc / image / URL / audio) which creates a notebook without leaving Flow. The name fits: a Notebook contains a flow of ideas; Flow turns it into a flow of frames.

Two ways in

Pick an existing NotebookLM — Flow lists the user's notebooks via the NotebookLM API, showing sources, sections and last-updated timestamps. Zero friction for the millions of users who already have notebooks.

Or attach sources inline — PDF, doc, image, URL, audio create an inline notebook without leaving Flow. NotebookLM API does the ingestion.

The brand fit

A NotebookLM contains a flow of ideas; Flow turns it into a flow of frames. Not a coincidence — a brand reinforcement.

Use cases observed in early discovery
  • Founders → pitch deck to explainer video
  • Teachers → lesson plan to kids' story
  • Authors → manuscript to book-trailer reel
  • Marketers → competitor analysis to pitch reel
  • HR → policy doc to onboarding video
  • Indian creators → Bhagavad Gita / Ramayana / textbooks to Shorts
4B · Plan & confirm — single credit-commitsingle credit commit
12:49
📶📡
83

Tara & the rain

Gemini plan
8 notebook sections → 6 shots · Gemini merged 3 short sections into the closing montage
edit
01
A rainy morning · Tara at the window
8s video9:1680 cr
02
Mother gives her a yellow raincoat
image9:1630 cr
03
Splashing in puddles, frog hops by
8s video9:1680 cr
04
She finds the rainbow · ending montage
8s video9:1680 cr
+ 2 more shots · tap to view all
smart transitions voiceover · Tara no title cards
Estimate ~3–4 min
4 video shots · Veo Fast320 cr
2 image shots · Imagen 460 cr
Voiceover · Tara (Gemini TTS)60 cr
Stitch + transitions20 cr
Total 460 credits
You have 3,840 cr · Flow Pro
2 free regenerations per shot included with this Flow — only the third onwards charges credits.
The single point of commitment, ElevenLabs-style. Gemini's proposed storyboard (compressed 8 notebook sections → 6 shots) is fully visible with per-shot model and credit cost. Options row covers stitching, voiceover, title cards. The estimate card breaks down every credit charge — no surprises. A 2-free-regenerations-per-shot guarantee is callout-pinned because that's the trust commitment that makes a "generate the whole thing" button safe to press.
4C · Review & regenerate — 2 free per shot2 free per shot
12:53
📶📡
82

Tara & the rain · v1

done
SynthID
stitched · 6 shots
0:58 · 9:16
Each shot
01
Rainy morning at the window
approved
02
Yellow raincoat handover
2 free regens left
03
Puddles & the frog
regenerating · 1 free left
04
Rainbow ending
free regens used · 80 cr to regenerate
After the batch generates, every shot is reviewable independently. The 2-free-regenerations-per-shot contract is enforced and visible: shot 02 has both regens unused, shot 03 is mid-regen with 1 free left, shot 04 has consumed both and now shows the credit cost honestly. Tweak ✎ opens that shot in the standard creator panel (Phone 4) so power-users can edit the prompt, change model, lock seed, etc. — and the regenerated shot drops back into the storyboard. Restitch & export rebuilds the final video at no extra cost — that's how iteration stays affordable.

The trust contract — why the user presses one button

The plan-and-confirm screen (4B) and the per-shot regenerate screen (4C) together form the trust contract that makes a "generate the whole thing" button safe to press. Four commitments:

  • Single commit, transparent estimate. The full breakdown — 4 video shots × 80 + 2 image shots × 30 + voiceover 60 + stitch 20 = 460 credits — is visible before anything generates. No mid-flow surprises. Estimate tolerance ±5 %, surplus refunded.
  • Two free regens per shot. ElevenLabs popularised this with voice. Generative AI is non-deterministic; charging full credits on every retry erodes trust. The counter is visible per shot: "2 free regens left" → "1 free left" → honest credit cost.
  • Restitch & export is free. Generation is the expensive part; stitching is cheap. Making it free removes friction on iteration.
  • "Tweak ✎" opens the standard creator panel (chapter 3, frame 3A). Power-users break out into the full playground for any one shot — prompt edit, model swap, seed lock, motion brush — and the result drops back into the storyboard.
  • SynthID + C2PA encode the source notebook ID in the provenance manifest. Verifiable origin — a moat no peer can claim.
  • Per-tier limits keep economics honest — Creator: up to 6 shots / 1 free regen per shot. Pro: up to 12 shots / 2 free regens per shot / Gemini Advanced bundled for ingestion. Studio: unlimited shots / brand kit auto-applied / provenance export.
  • This is glue, not a silo. Notebook flow uses Storyboard mode under the hood, opens shots in the standard creator panel, deposits assets in the Library as a smart collection, and routes the final stitch to YouTube Shorts (chapter 5).
Chapter 05 · 5 frames

Google handoffs: keep what only Google can do, gate the rest.

The Beta force-bundles every Google product as if all are equally relevant. v1.0 separates them by whether the handoff belongs in Flow's headline experience: YouTube Shorts publish stays in the front door (Google owns it), storage / Music / Gemini Advanced move to progressive disclosure, and Calendar / Gmail / Tasks drop out.

The handoff verdict matrix

Google productVerdictSurface in Flow
YouTube Shorts publishKEEP · headlineResult-screen action row + full publish flow (5A)
NotebookLM ingestionKEEP · flagship workflow"From a Notebook" hero (chapter 4)
Google Photos readKEEP · existingAsset picker (chapter 3)
SynthID + C2PA provenanceKEEP · always-on defaultBadge on every output + controls (chapter 6)
Voice prompt via Gemini stackKEEP · free, on-device where possibleVoice input on creator (5D)
Imagen 4 · Nano Banana Pro · Veo 3.1KEEP · core modelsThese are Flow's models, not Google One features
Google One storageOPT-IN · progressive disclosureSurfaces at 96 % asset space (5B)
Flow Music studioOPT-IN · progressive disclosureSurfaces when adding audio (5C)
Gemini Advanced + NotebookLMOPT-IN · bundled at Pro for Notebook flowManage Flow add-on row (5E) for non-Pro
Google Drive exportOPT-IN · mutedGray chip in post-actions row
Workspace SSOOPT-IN · Studio onlyStudio tier add-on
Google Calendar / Gmail / TasksOUTNot relevant to the generation persona
5A · Publish to YouTube Shortsnative handoff
12:45
📶📡
84

Publish to YouTube

Mumbai monsoon · v3
0:08 · 9:16
SynthID
Title
Mumbai monsoon vibes 🌧️ #shorts
Description
Cinematic dolly-in over a vintage red car · made with Flow + Veo 3.1
#aigenerated #flow #monsoon
Public
For kids · No
AI-generated content disclosed
YouTube auto-applies the "altered or synthetic" label. SynthID + C2PA travel with the file.
The one Google-suite integration that belongs in the headline flow. YouTube Shorts is the obvious destination for 9:16 generated video, and Google owns it — so we keep the publish inside Flow with title, description, hashtags, visibility, kid-safety toggle, and the auto-applied AI-disclosure label pre-set. SynthID + C2PA piggy-back. Higgsfield is the only peer with anything comparable; Sora's "publish" is to its own feed, which Flow deliberately doesn't have.

Why YouTube Shorts is kept in the headline flow

  • 9:16 is already the dominant generation aspect ratio — matches Sora's defaults and Veo's preset
  • One-tap upload with title / description / hashtags / visibility / kid-safety pre-filled is a meaningful time-saver
  • Google's ad ecosystem benefits — Flow becomes top-of-funnel for Shorts
  • Provenance (SynthID + C2PA) travels with the file; YouTube's AI-generated content label is auto-applied
  • Only Higgsfield among peers has anything comparable; Sora's "publish" is to its own feed, which Flow deliberately won't have
Target KPI

YouTube Shorts publish rate per 9:16 generation: 18 % at day 90.

5B · Storage cap — Google One opt-inprogressive disclosure
12:38
📶📡
85
Google Flow
BETA
D
Generate
image
Generate
video
Animate
Edit

Your Flow space is almost full

You're using 4.8 GB of 5 GB. Free up space — or extend your storage with Google One, only if you'd like to keep everything.

4.8 GB used5.0 GB Flow Creator
Google One · 2 TB For heavy export workflows
+₹650/mo
The first progressive disclosure moment. Storage is the one place where Google One genuinely helps a Flow user — so we surface it only here, in context, with clear ₹149/mo and ₹650/mo prices that ride on top of the Flow plan. Compare with today: storage is force-bundled inside the AI Plus / AI Pro / AI Ultra headline price, whether you need it or not.
5C · Add music — Flow Music studio opt-inprogressive disclosure
12:39
📶📡
85

Add sound

No audio
Choose a sound source
Upload an audio clip, pick from your library, or generate something new.
Upload audio file
Royalty-free library
2,400 tracks

Flow Music studio ADD-ON

Generate an original score from a text prompt or hum a melody. Includes SFX foley library and stem-level export.

+₹199/mo on top of your Flow Pro plan · cancel anytime
Surfaced because you're trying to add audio. We never bundle Music into Flow Pro by default — only when you actually open this flow.
The second disclosure moment. Today, "Flow Music studio" is force-bundled into Google AI Plus (₹399), so non-musical users effectively subsidize it. Here it appears only when the user tries to add audio — an honest cross-sell. Upload & royalty-free library remain free.
5D · Voice prompt → Gemini storyboardprogressive disclosure
12:47
📶📡
84

New idea

listening
You said
"Banao ek 30-second ad, monsoon mein vintage red car ke saath, Aarav ke saath jo lights cigarette outside cafe…"
🇮🇳 Hinglish 12 s recording
Gemini turned that into a 4-shot storyboard ADD-ON
+₹599/mo unlocks prompt-assist, scene expansion and 100+ daily Gemini calls.
01
Wide est. shot
02
Aarav reveal
03
Match cut
04
Pull-back
Or just use what you said
Generate 1 video from this prompt
20 cr
Two Google-only powers, both progressively disclosed. Voice prompt in Hinglish (Gemini speech stack, on by default on supported Pixel and Android phones — free) and Gemini storyboard expansion which turns one sentence into a 4-shot breakdown (opt-in via Gemini Advanced add-on, ₹599/mo). User can always skip and just generate from the raw transcription — no Gemini gate required.

Three progressive-disclosure moments

5B · Storage cap — surfaces only at 96 % asset space. Google One 200 GB (+₹149) or 2 TB (+₹650) sit honestly inside a Flow-branded sheet. No more force-bundling.

5C · Music studio — surfaces only when the user opens the audio flow with no track uploaded. Flow Music studio at +₹199/mo, opt-in.

5D · Voice prompt → Gemini storyboard — voice input itself is free (Gemini speech stack, on-device on Pixel), Hinglish supported. The storyboard expansion is the Gemini Advanced add-on (+₹599/mo for non-Pro) — user can always skip and just generate from the raw transcription.

Each cross-sell appears at the boundary where it solves a real problem. Bundle by capability, never by SKU.

Manage Flow — add-ons as honest line items

Replaces today's "Manage subscription" → Google One handoff. Everything Flow lives inside Flow: credit meter, attached add-ons (each priced and removable), and an unbundled upsell to Gemini Advanced + NotebookLM as a separate plan. Cancel Flow is one button, not a Google One cancellation that strips your storage too.

What disappears from this screen vs Beta
  • Routing to Google One billing surface
  • Forced bundle of storage / Gmail-AI / NotebookLM
  • Plans named after "AI Plus / Pro / Ultra"
5E · Manage Flow — add-ons as line itemsno Google One handoff
12:41
📶📡
85

Manage Flow

Flow Pro
Renews June 24, 2026 · ₹1,499/month
4,210 / 6,000 credits resets Jun 24
≈ 9 days of usage left at your current rate
Active add-ons
Google One · 200 GB
Added May 19 · ₹149/mo
Remove
Not added
Flow Music studio
+₹199/mo · AI score + SFX
Add
Gemini Advanced + NotebookLM
+₹599/mo · upgrade to full Google AI
Add
Replaces today's "Manage subscription" → Google One handoff. Everything Flow lives inside Flow: credit meter, attached add-ons (each priced and removable), and an unbundled upsell to Gemini + NotebookLM as a separate plan. Cancel Flow is a single button, not a Google One cancellation.
Chapter 06 · 2 frames

Trust & safety: the moat no peer can match cheaply.

SynthID is Google DeepMind's invisible pixel-watermark; C2PA is the industry-standard cryptographic manifest. Both ship on by default in Flow, can't be turned off, and travel with every export — including YouTube. This is the brand position no peer can credibly take.
6A · Provenance & safety — SynthID + C2PA always onalways on
12:46
📶📡
84

Trust & provenance

Everything you make ships with proof it's AI.

SynthID watermark and C2PA content credentials are baked into every Flow output. You can't remove them — and that's the point.

SynthID — invisible watermark
Pixel-level, survives crop & recompress.
ON
C2PA Content Credentials
Cryptographic manifest of model, edits, timestamp.
ON
Controls
Verify a Flow asset
Visible watermark on Free tier
Block deepfake of public figures
Flow refuses prompts targeting named real people without verified consent. Cameos require face verification. Reporting tool surfaces in every share sheet.
The defensible Google moat almost nobody else can match cheaply. SynthID is Google DeepMind's invisible pixel-watermark; C2PA is the industry standard manifest. Both ship on by default, can't be turned off, travel with every export (including YouTube). Flow as the trust-first generation product is the right brand position for Google — and the right answer to growing deepfake regulation.

Provenance is platform, not feature

SynthID survives crop, recompress, and screenshot. C2PA encodes the model, edits and timestamp cryptographically. Together they answer: was this made with AI? when? by whom? from what source?

Controls (not removable, but configurable)
  • Verify a Flow asset — paste any image/video, check its provenance
  • Visible watermark on Free tier — toggleable on paid tiers
  • Block deepfake of public figures — refuses prompts targeting named real people without verified consent
Where this also shows up

The Notebook flow (chapter 4) encodes the source notebook ID into the C2PA manifest, so the chain is verifiable from grounded source to final video. Public verification tool ships Phase 5.

Cameos with face verification — no public registry

Flow borrows Sora 2's face-verification step (turn-head clip) so likeness can't be stolen. But unlike Sora, Flow refuses to host "public" cameos at all. Three permission tiers:

  • Only me ✓ (default)
  • Studio teammates — explicit add
  • Public — disabled by policy. Flow does not host a public cameo registry.
Why this matters

Sora's open cameo registry triggered an abuse cycle within weeks of launch. Google can't ship that. The 3-tier model is the safety prerequisite for a serious character system.

6B · Train a character — face-verified, no public cameosface-verified
12:44
📶📡
84

New character

Teach Flow what Aarav looks like

Upload 8–20 photos in varied poses, lighting and angles. Take a short verification clip so we know the likeness is yours to use.

Photos · 14 / 20
good coverage
+
Verification recorded
3-second clip of you turning your head. Used to prove this likeness is yours — never shared, never used to train base models.
Who can use Aarav
Only me
Studio teammates
2 collaborators
Public
Disabled — Flow does not host public cameos
The Character library entry-point. Borrows Sora 2's face-verification step (turn-head clip) so likeness can't be stolen — but unlike Sora, Flow refuses to host "public" cameos. Three permission tiers: Only me, Studio teammates, and Public (disabled by policy). This is the safety prerequisite for a serious cameo / character system.

Anti-features: what Flow deliberately won't ship

  • TikTok-style social feed — Sora 2 proved it's brittle; Google has YouTube Shorts
  • Full NLE timeline — Runway tried; users export to Premiere / Resolve / CapCut
  • Third-party model aggregator (Krea-style 64+ models) — Google's promise is integrated quality, not breadth
  • Public character cameos — abuse vector
  • Auto-publish anywhere — provenance + opt-in publish only
  • Removable provenance / watermark — SynthID + C2PA can't be turned off
  • Credit-price-prominent Free UX — show "1 generation left today," not "20 credits"
  • Hidden mid-flow charges in Notebook batches — single commit, free regens, restitch free — or the trust contract breaks
Chapter 07 · 2 frames

Pricing: standalone tiers, competitive band, no storage tax.

Four standalone tiers — Free / Creator / Pro / Studio — priced inside the $5–$30 competitive band the rest of the category lives in. Storage, Music and Gemini Advanced become opt-in add-ons surfaced at the boundary where they solve a real problem.

The standalone tier set

TierIndia ₹US $CreditsModelsCreator toolsFrom a NotebookAsset space
Free₹0$040 / day Veo Lite, Imagen 4, NB 2 Prompt, ratio, model, voice prompt Preview only 1 GB · watermark
Creator₹399$4.991,500 + Veo Fast, NB Pro + Motion brush, Keyframes, Style, Refs, Seed, Negative, Image playground up to 6 shots · 1 free regen / shot 5 GB · no watermark · commercial
Pro ★₹1,499$17.996,000 All incl. Veo Quality, Omni Flash + Camera, Character library, Storyboard, Lip-sync, Extend, Post-gen Aleph toolbar up to 12 shots · 2 free regens / shot · Gemini Advanced bundled 25 GB · 4K upscale · priority queue
Studio₹4,999 / seat$59.99 / seat25,000 pooled All + API All + brand kits, API beta, SSO unlimited shots · brand kit auto-applied · provenance export 250 GB · team

Add-ons (any paid tier): Google One storage (200 GB +₹149 / 2 TB +₹650) · Flow Music studio (+₹199) · Gemini Advanced + NotebookLM standalone (+₹599) — for Creator who wants deeper Notebook usage outside Flow.

7A · Pricing redo — standalone Free / Creator / Pro / Studiostandalone Flow plans
12:36
📶📡
86

Flow

Plans built for image & video creators.

Credits, models, and creator features only. Add cloud storage, music studio, or Gemini upgrades only if you actually need them — never bundled by default.

Save 18% when you pay annually
Flow Free
Try Veo 3.1 Lite, Imagen 4 and Nano Banana Pro every day.
₹0 /forever
  • 40 daily credits, refresh on first use
  • Veo 3.1 Lite · Imagen 4 · Nano Banana 2
  • 5 saved projects · 1 GB Flow asset space
  • Personal use only · subtle watermark
Flow Creator
For Reels, Shorts, ad drafts and weekend films.
₹399 /month
  • 1,500 monthly Flow credits
  • Veo 3.1 Fast · Nano Banana Pro · Imagen 4
  • Up to 10s clips · 1080p · no watermark · commercial license
  • 5 GB asset space · unlimited projects
Flow Studio
Teams, agencies, high-volume production.
₹4,999 /month / seat
  • 25,000 monthly Flow credits / seat · pooled
  • Shared brand kits, characters, voices
  • Team library · approvals · usage rollups
  • 250 GB asset space · API access (beta)

Need more? (optional, never bundled)

Surface only when your usage suggests it — e.g. asset space nearly full triggers the storage card; importing audio triggers the music studio card.

Google One · 200 GB cloud
Back up assets across devices. Pay only if Flow space fills.
Flow Music studio
AI score + SFX library. Add to any plan for ₹199/mo.
Gemini Advanced + NotebookLM
For when you outgrow Flow alone. Cross-suite plan.
Cancel anytime. Prices shown for India. By subscribing you agree to the Flow terms, separate from Google One.
The screen the user asked us to build. Four Flow-only tiers (Free / Creator / Pro / Studio). No storage forced into the headline price. Google One, Music studio and NotebookLM live below as optional add-ons — visible but explicitly opt-in. Compare with the current Beta routing users into Google One AI Plus → AI Pro → AI Ultra.

Indian ₹ band rationale

  • Creator at ₹399 matches today's AI Plus headline price — but the user keeps the money that currently funds 200 GB of unused storage
  • Pro at ₹1,499 is 23 % cheaper than AI Pro (₹1,950), and bundles Gemini Advanced for the Notebook flow — a meaningfully better value
  • Studio at ₹4,999 undercuts AI Ultra (₹6,500) and adds the team / API features Ultra doesn't have
Cohort math

1,000 generation-active free users at the upgrade decision:

  • Beta: 17 convert to AI Pro at $19.99 → $339 MRR
  • v1.0: 23 convert at avg $14 (mix of Creator/Pro) → $322 MRR
  • + 6 attach Google One 200 GB at +₹149 → $333 MRR
  • + 2 attach Music, 1 attaches Gemini Advanced → $343 MRR
  • Storage-adders now pay less than under bundled AI Pro → churn drops ~40 % → LTV +35 %

Plus a new acquisition funnel: NotebookLM cross-links bring in users who would never have hit the prompt-driven funnel (teachers, founders, authors). +8 % new accounts, converting to Pro at ~4–5 % rates because Notebook flow is the headline feature they want.

Account screen — minimal redesign, maximum impact

Layout untouched from Beta — same Darshak avatar, same "40 Flow credits" indicator, same My library / Help / Manage / Sign out rows. The single change: "Upgrade Flow" now routes to the standalone Flow pricing screen (chapter 7A), not to a Google One bundle picker.

"Manage subscription" is renamed "Manage Flow membership" to set the right expectation (chapter 5, 5E).

7B · Account — Upgrade routes to Flow, not Google Oneexisting layout
12:36
📶📡
86

Account

D
Darshak darshakbhatt1994@gmail.com
40 Flow credits
Credit refreshes daily · You're on Free
Layout untouched — but "Upgrade" now points to a standalone Flow pricing screen, not Google One. "Manage membership" renamed "Manage Flow membership" to set expectations.
Chapter 08

Impact & rollout: 17 KPIs, 5 phases through 2027 Q2.

KPIs to instrument before launch, with 90-day targets. Rollout phased so Pro tier is shippable at Phase 1 without waiting for every advanced tool. Notebook flow goes beta at Phase 3 (2026 Q4 → 2027 Q1) and GA at Phase 4.

Day-90 KPI targets

MetricBeta baselineDay 90 targetLever
Activation (% new accounts generating ≥1 asset, wk 1)41 % est.62 %8-mode home + Notebook entry + voice prompt
Paid conversion (free → paid, 30 d)1.7 %2.3 % (+38 %)Standalone ₹399, no storage tax
Blended ARPU (India)₹1,460₹1,780 (+22 %)Pro anchors better; add-on attach
Add-on attach · storagen/a (forced)28 % of Creator+Triggered at 96 % usage
Add-on attach · Musicn/a9 % of ProTriggered only when adding audio
Add-on attach · Gemini Advanced (Creator)n/a12 %Voice prompt & Notebook depth triggers
Notebook → Flow attach (Pro users, 30 d)n/a (new)35 %Hero card + existing-notebooks picker
Notebook flow completion (plan → stitch)n/a78 %Single commit, free regens, restitch free
Avg regens per shot in Notebook flowsn/a1.4Within 2-free allowance — Gemini sizing
Motion brush sessions / wk per Pro user04.2First-class in creator panel
Multi-shot storyboards / mo per Pro user01.8Storyboard mode + character continuity
Post-gen edits per generation00.7Aleph toolbar (chapter 3)
Library DAU among users w/ ≥30 assetslow+60 %Search, tags, smart collections
YouTube Shorts publish rate per 9:16n/a18 %One-tap publish + metadata pre-fill
30-day paid retention64 %74 %Lower regret; trust contracts
NPS (post-upgrade)+12+28Paying for what you actually use
Churn citing "paying for storage I don't use"18 %< 5 %Removed from headline
Acquisition · % new accts via NotebookLM cross-link08 %New funnel: NotebookLM users discovering Flow

Rollout — 5 phases

PhaseScopeTiming
Phase 0 · InternalStandalone billing plumbing; SKU split from Google One2026 Q3
Phase 1 · India + SEA pilotStandalone Free / Creator / Pro; Library with search & tags; YouTube Shorts publish; SynthID + C2PA default2026 Q3 end
Phase 2 · Playground depthMotion brush, Keyframes, Camera, Refs, Seed lock, Negative, Character library w/ face verification2026 Q4
Phase 3 · Post-gen + Notebook betaAleph-class toolbar, Image playground (inpaint / outpaint / remix), 4K upscale, Voice prompt, Notebook flow beta at Pro2026 Q4 → 2027 Q1
Phase 4 · Studio + agents GAStudio plan, brand kits, API beta, pooled credits, Notebook flow GA with Creator-tier access, real-time canvas (Pro+), additional workflow agents2027 Q1
Phase 5 · Global GARoll standalone to all markets; legacy AI Pro migration; Beta tag drops; provenance verification tool public; NotebookLM ↔ Flow deep cross-links2027 Q2

Risks & mitigations

RiskMitigation
Cannibalises Google AI Ultra revenueMost Ultra subs buy for Deep Think / Mariner / NotebookLM standalone. Track migration; expect ≤ 8 % downgrade.
Storage add-on attach lower than 28 %Even 10 % attach is incremental margin vs today's waste.
Feature parity ambition outruns engineeringPhased rollout. Pro shippable at Phase 1 without Motion brush etc.
YouTube publish surfaces brand-safety issuesAI-disclosure label non-removable; SynthID + C2PA travel with file; reporting tool in every share sheet.
Character library abused for deepfakesFace-verification mandatory; Public cameos disabled by policy.
Notebook flow credit estimate wrongEstimate built from deterministic model-pricing math; tolerance ±5 %, surplus refunded.
User burns free regens, feels punishedShow cost honestly upfront; Restitch free; "Tweak" reopens full playground.
NotebookLM compute costs at scaleGemini Advanced bundled at Pro; Creator users get rate-limited preview only.
Chapter 09

Scalability: six axes Flow is architecting for.

Not all of these ship at launch, but the architecture must support them. The Notebook flow (chapter 4) is the first concrete instance of axis 3 — the rest fall into place behind it.

1 · Multi-model routing

Each shot in a storyboard can route to a different engine — Veo Lite for drafts, Veo Quality for hero shots, Imagen for stills. Pro+ shows per-shot cost preview. Higgsfield's approach minus the third-party aggregation; Flow stays Google-models-first.

2 · Real-time canvas

Krea's sub-50 ms regeneration is the new bar for image work. Pro+ "Live preview" toggle on the image playground (chapter 3, 3B) once GPU pipelines support it. Path: Imagen 4 fast-path → on-device draft using Gemini Nano on Pixel for the lightest pass.

3 · Workflow agents

From a Notebook (chapter 4) is the first shipped agent. Phase 4 adds: marketing agent ("4 ad variants for different age groups, render 9:16, publish to YouTube Shorts with A/B metadata"), sales agent ("turn the last 6 customer-call transcripts into a 60s product reel"), publishing agent ("12-shot weekly newsletter video from my latest blog post"). All reuse the plan / commit / regen pattern.

4 · Region-aware models

Indic LoRAs on Imagen / Veo for Indian faces, fashion, architecture, regional vehicles. Hinglish prompt understanding (chapter 5, 5D). A hard moat — Western model labs won't ship this fast, and India is one of Flow's anchor markets.

5 · Provenance as platform

SynthID + C2PA on every output (chapter 6) is the start. Future: shared C2PA manifest format with YouTube, Photos and Drive so the chain is verifiable across the Google ecosystem. For Notebook flows, the manifest encodes the source notebook ID — a verifiable origin claim. Public verification tool ships Phase 5.

6 · Trust as differentiator

Three things make Flow's promise different from competitors': (a) provenance default-on, (b) single credit-commit with free regens (chapter 4), (c) face-verification for cameos with no public registry (chapter 6). Together they form a brand position no peer can credibly take — and one Google specifically needs to occupy.

Chapter 10

Playground maturity matrix — Flow v1.0 vs every serious peer.

17 capabilities scored ● Full / ◐ Partial / ○ Missing. Two rows isolate the moat — Source-grounded ingestion (NotebookLM) and Batch commit + free regens — where every other peer scores zero.

Capability comparison · verified May 2026

Capability Flow Beta Flow v1.0 Runway G4 Kling 3.0 Pika 2.5 Sora 2 Grok Im. Krea
Text→Image
Image→Image edit · multi-ref
Motion brush
Keyframes / start-end
Multi-shot stitching
Character library + cameo
Lip-sync · native audio
Post-gen edit (Aleph-class)
Extend / re-cut / remix
4K upscale
Seed lock · negative prompt
Library: search + tags
Voice prompt
Direct publish to social● YT Shorts◐ Feed◐ X
Provenance: SynthID + C2PA● default
Source-grounded ingestion (PDF, doc, audio → storyboard)● NotebookLM
Batch commit + free regens● 2/shot
Full · Partial · Missing Sources: Kling 3.0 (Kuaishou), Runway Gen-4 (runwayml.com), Pika 2.5 (pika.art), Sora 2 (OpenAI; web/app shutdown 26 Apr 2026), Grok Imagine (x.ai), Krea (krea.ai). Verified May 2026.

Built by

Darshak Bhatt
Co-creator · Claude (Anthropic)
Back to overview