SiteAsk AI
CV (PDF)
All work

Pitchy

A two-way text to speech editor for video voice-overs.

Role
Freelance product engineer
Period
Jan – Apr 2021
Sector
Marketing video creation, France
Pitchy interface
  1. 01Still in production, five years onShipped in 2021, the feature is still part of the product, now with an AI voice engine
  2. 02One editor, two ways to writeA visual editor for most users and raw SSML for power users, kept in sync
  3. 03Built end to end in 4 monthsEditor, interface design and backend, as a freelancer working with the product owner

Pitchy is a platform for creating marketing videos.

Users who wanted a voice-over had to record it themselves. Pitchy wanted them to type a script and get a generated voice instead.

I joined as a freelancer for four months to build this feature, working with the product owner and handling the editor, its interface design and the backend.

The goal was to let anyone add a voice-over to a video without recording one.

Team
Me, with a product owner · tech team of ~10
Stack
React · Node.js / Express
Voice
AWS Polly · SSML

Three problems to solve before a script could become a voice-over.

  1. Problem 01

    Controlling a generated voice (pauses, emphasis, pronunciation) is done in SSML, a markup language most users would never write.

    Solution

    A visual editor that produces SSML behind the scenes.

  2. Problem 02

    Power users wanted to write SSML directly, without losing what the visual editor offered.

    Solution

    A two-way editor: changes on either side update the other.

  3. Problem 03

    The voice had to match what happens in the video.

    Solution

    Pauses in the text, and the ability to start a sentence at a precise timecode.

How we built it, step by step.

Framing with the product owner

We started from the path a script would follow: typed by the user, shaped in the editor, turned into SSML, voiced by AWS Polly and placed as an audio track in the video's timeline.

UserScript
Two-wayEditor
MarkupSSML
VoiceAWS Polly
VideoTimeline

Still in production, five years on.

Shipped in 2021, the feature is still part of Pitchy today, now running on an AI voice engine.

One editor, two ways to write.

Most users write visually, power users write SSML, and both stay in sync.

Built end to end in 4 months.

Editor, interface and backend, delivered as a freelancer.

Next case study

MadeForMed — 2026Merging two medical products into one, starting with the interface.