ECHO — Live
A voice note in. A fortnight of content out.
ECHO is a content repurposing platform built from nothing — strategy, product design, brand, architecture and code. It turns a founder’s voice note into platform-ready posts written in their own voice, ready to approve and schedule.
Figures are pre-launch. ECHO opens to the public on 8 September 2026 — refresh these after the switch.
What it is
ECHO is software for people who have plenty to say and no time to write it down eight different ways. You record a voice note — the thing you’d say to a client, or the thought you had in the car — and ECHO turns it into a set of finished posts, one for each platform you actually use, written the way you write.
It is not a prompt box with a logo on it. Before ECHO writes a word, it takes a founder through a structured onboarding that captures how they actually sound: the vocabulary, the sentence rhythm, the things they refuse to say. That profile is the product. Everything generated afterwards is measured against it.
Then the ordinary work of a content tool: a queue to review in, inline editing, regeneration where something missed, and export or scheduling once it’s right. Nothing goes out that a human hasn’t approved.
Why it was built
Twenty years of marketing taught the same lesson every time: the bottleneck was never ideas. It was the hour it takes to turn one good thought into a fortnight of posts, and the fact that hour never happens.
Generic AI didn’t solve it. It produced content that was technically correct and unmistakably nobody’s — the kind of writing an audience scrolls past because it doesn’t sound like the person they followed.
The missing piece wasn’t a better model. It was the brand voice framework already being used with clients by hand. Turn that into onboarding, feed it to the model as the brief, and the output stops sounding like AI and starts sounding like the founder. That is the whole product, and it is the reason ECHO exists rather than a wrapper.
What’s inside
The product itself.
Drag or scroll Swipe
Three ways in — type or paste, record, or upload a voice note or video. Everything downstream is identical, so the founder picks whichever is nearest to hand. The box underneath steers the angle: what to build from, what to skip, what to close on.
One input, every output, generated in parallel rather than one at a time — so the wait is one wait, not eight. Twenty to forty seconds, and it says so rather than spinning silently.
Ten outputs from one voice note, each written for the platform it is going to. Read them, deselect what missed, send the rest to drafts. Nothing publishes itself.
The help centre, and the door to MAVIS — the assistant that answers from it.
How it was built
Strategy, product, build, launch.
Strategy
Start with the buyer, not the feature list. Who is stuck, what they’ve already tried, what they’d pay to stop doing. The answer set the whole scope: one job, done properly, rather than a content suite nobody finishes setting up.
Product
A written specification before a line of code — the ten onboarding steps, the three ways content gets in, the approval loop, and an explicit out-of-scope list. Image generation and video were named as later phases so the first release could actually ship.
Build
React and TypeScript on the front, Node and Express behind it, Supabase for database, auth and file storage, and the Anthropic API doing the generation against a system prompt assembled from the user’s own brand profile. Transcription handles the voice notes.
Launch
A private beta first, with real founders on three continents using it on their own content and saying what was wrong. Then a help centre, an in-app assistant, billing, and a founding cohort — before the public door opens on 8 September 2026.
The outcome
What it did.
ECHO’s founding round sold out in seven hours and took £17,500, before the product was publicly available — from a cohort spread across three continents who were using it on their own businesses rather than testing it as a favour.
It is a working SaaS product with authentication, a database, file storage, a generation engine, a billing system, a help centre and its own in-app assistant — built and shipped in-house rather than commissioned. It is the reason this studio can say it builds software instead of saying it manages people who do.
Questions
What people ask about ECHO.
Can a developer build an AI app that turns voice notes into content?
Yes — that is exactly what ECHO is, and it was built from nothing. The pipeline is: capture the audio, transcribe it, hold a structured profile of how the user writes, then generate each platform’s output against that profile rather than against a generic prompt.
The hard part is not the transcription or the API call. It is the brand voice layer in between, which is what stops the output reading like every other AI post on the timeline.
How do you stop AI-generated content sounding like AI?
By giving the model a real brief instead of a vague instruction. ECHO’s onboarding captures vocabulary, sentence rhythm, the words a founder refuses to use, their audience, and three to five samples of their actual writing — then every generation is measured against that.
The samples matter most. Showing a model how someone writes works far better than describing it.
How long does it take to build a SaaS product like this?
ECHO went from specification to a paying private beta inside a few months, then spent longer being hardened for public launch than it did being built. That is the usual shape: the first working version arrives fast, and the distance between working and sellable is where most projects quietly die.
A smaller product with one job and one screen is a matter of weeks. Something with accounts, billing and a generation engine is months. Anyone quoting you a fortnight for the second one is quoting for the first.
Can I have a SaaS product built for my own methodology?
That is the same job as ECHO. ECHO productised a brand voice framework that previously only existed as a consultant working through it with a client by hand. Any methodology with steps, judgements and a repeatable output can be built the same way.
It is the work described on the custom app development page, and it is what most of this studio’s enquiries actually turn out to be.