AI assistants and search decision
Speaky
A technical user can build a usable live speech→slides replacement covering Freestyle and basic Guided modes, but reproducing the vendor's low-latency, highly polished experience (and ongoing product polish) is costly and operationally heavier than a small self-hosted project.
Visit website↗Built by Alberto Ravasini ⚡, who ships 5 products in this index
$5/mo
$60/yr
Read off the official pricing page.
$100one-off80 h to build
$150/mo6 h/mo upkeep
On cash alone, building overtakes the subscription at 32 seats.
No open-source build does this yet
Nothing published replaces this one, so a replacement starts from an empty file. Here is what it would have to cover.
What a replacement has to do
- Speak → realtime speech-to-text → intent/layout extraction → render slide visuals/overlays → save transcript and slide history
What it still won’t have
- Sub-second, Cerebras-backed inference speed and the claim of slides appearing "before you finish your sentence"
- Polished commercial UX, weekly product updates, and the vendor's support/guarantees
- Any proprietary optimizations for instant overlays and trigger-word responsiveness
What remains hard
- Product polish and ongoing maintenance
First-year cost
Keep paying
Paying is—cheaper in year one.
On cash alone, building overtakes the subscription at 32 seats.
Money you would actually spend
Time you would spend
—
What you would spend
What we assumed
The verdict above measures whether you could build it. This one is only about money.
Runnable build prompt
Build a self-hosted minimal live voice-to-slides service using Next.js + React for the frontend, a Node.js/Express WebSocket server, Postgres for storage, and Redis for transient state. Integrate a cloud ASR (e.g., Whisper/any streaming ASR API) for real-time transcript and OpenAI/GPT-style model calls (or local LLM) for layout intent extraction. Core features in scope: live speech streaming with transcript, mapping utterances to 8 layout types, server-side render of slide frames (HTML/SVG) and instant numeric overlays, upload and triggerable mockups, manual slide editor and save/playback of presentations, basic email auth and per-user storage. Out of scope: training custom models, matching Cerebras-level sub-second latency optimizations, multi-tenant billing, analytics dashboard, and mobile app. Include error handling for ASR/model failures, tests for the WebSocket endpoints and core intent extraction, and Docker Compose for local deployment.
How we checked
How the score was reached
- Partly verdict base52
- 3 cited sources+3
- Price verified on pricing page+3
- 3/3 assessment runs agreed+4
- Evidence score62
The base comes from the verdict. Everything under it is a check that either happened or did not, and each one is a fact frozen in this record rather than a judgement made at render time - so the same evidence always produces the same number.
How scoring works →Cited sources · 3
Every page the run actually retrieved.
- official productSpeaky.st — Live Presentations — No Slides Required
- official pricingSpeaky.st Pricing — Go Live $5/mo
- official productSpeaky.st — Sub-second AI claim
Integrity checks
What held up, and what did not.
