EdgeSpeak is LattifAI's desktop app for local speech work. It runs the Lattice-2 model on your computer for transcription, forced alignment, speaker diarization and translation, and exposes the same capabilities through a local API, CLI and MCP.

Where did the online alignment and transcription tools go?

They have moved into EdgeSpeak. The desktop app fully supports forced alignment and transcription with Lattice-2, and processes your media locally instead of in the browser or the cloud.

Is my audio uploaded anywhere?

No. Transcription, alignment and diarization run on your device, and your audio and text are not uploaded to a third-party cloud. License activation and model downloads do require a network connection.

Which platforms are supported?

The desktop app runs on macOS 14+ (Apple Silicon) and Windows 10/11 x64. The CLI is also available on Linux x86_64.

How much does EdgeSpeak cost?

EdgeSpeak is a one-time purchase with no subscription and no per-minute fees. Each device gets a 7-day free trial, and one license covers up to 4 devices. See edgespeak.com for current pricing.

NEW

Lattice-2 is now live in EdgeSpeak

Lattice-2, now in EdgeSpeak

LattifAI's Lattice-2 model powers EdgeSpeak: transcription, forced alignment, speaker diarization and translation, all running on your own computer.

Download EdgeSpeak

Runs on your device

•

No audio uploaded

•

Word-level timestamps

Our flagship product

EdgeSpeak: Lattice-2 on your own device

Forced alignment, transcription, speaker diarization and translation now run fully in EdgeSpeak — no audio or text uploaded to the cloud.

Learn more at edgespeak.com

Lattice-2 alignment & transcription, 40+ languages
Runs locally on macOS & Windows
Desktop app, local API, CLI & MCP

CORE FEATURES

Lattice-2 inside EdgeSpeak

Everything the online tools did, now on your desktop

EdgeSpeak brings the Lattice-2 engine to macOS and Windows, with a desktop app, local API, CLI and MCP.

Local Transcription

Turn audio and video into text on your own computer.

40+ Languages

Lattice-2 Flash / Pro supports more than 40 languages.

Word-level Timestamps

Export transcripts and captions with accurate timing for every word.

Local Transcription - LattifAI feature demonstration: Turn audio and video into text on your own computer. (Image 1 of 2)

Local Transcription - LattifAI feature demonstration: Turn audio and video into text on your own computer. (Image 2 of 2)

Forced Alignment & Speakers

Align an existing transcript to its audio and see who said what.

Forced Alignment

Already have the script or lyrics? Get precise start and end times for each word.

Speaker Diarization

Separate speakers in interviews, meetings and podcasts.

Translation

Translate transcripts while keeping every timestamp intact.

Forced Alignment & Speakers - LattifAI feature demonstration: Align an existing transcript to its audio and see who said what. (Image 1 of 2)

Forced Alignment & Speakers - LattifAI feature demonstration: Align an existing transcript to its audio and see who said what. (Image 2 of 2)

Private by Design

Your audio and text stay on your device.

No Uploads

Media is processed locally and is never sent to a third-party cloud.

Built for Workflows

Use the desktop app directly, or connect apps and agents through the local API, CLI and MCP.

Honest Boundaries

License activation and model downloads need a network connection; speed depends on your hardware.

Private by Design - LattifAI feature demonstration: Your audio and text stay on your device.

COMPARISON

EdgeSpeak vs cloud transcription

Why run it locally?

Local isn't always better than the cloud. It's the right choice when privacy, data control and long-term cost matter most.

Aspect	Cloud services / APIs	EdgeSpeak
Data path	Audio or text is sent to the provider's servers	Processed on your device, nothing uploaded
Cost	Usually billed per minute, character or call	One-time purchase, no subscription, no per-minute fees
Control	Provider controls model versions, regions and limits	You choose device, models and entry points
Product form	Mostly web apps or remote APIs	Desktop app, local API, CLI and MCP
Scale	Better for elastic, high-concurrency workloads	Bounded by your hardware

Data path

Cloud services / APIs:

Audio or text is sent to the provider's servers

EdgeSpeak:

Processed on your device, nothing uploaded

Cost

Cloud services / APIs:

Usually billed per minute, character or call

EdgeSpeak:

One-time purchase, no subscription, no per-minute fees

Control

Cloud services / APIs:

Provider controls model versions, regions and limits

EdgeSpeak:

You choose device, models and entry points

Product form

Cloud services / APIs:

Mostly web apps or remote APIs

EdgeSpeak:

Desktop app, local API, CLI and MCP

Scale

Cloud services / APIs:

Better for elastic, high-concurrency workloads

EdgeSpeak:

Bounded by your hardware

Your media, your machine

EdgeSpeak is for people who would rather not upload every recording to the cloud.

USE CASES

What people use EdgeSpeak for

Built for real workflows

From captions to datasets, processed on your own computer.

Captions & Video Editing
Generate word-timed captions and align scripts for faster post-production.
Speech Datasets
Batch-transcribe and align audio on your own machines with the CLI.
Meetings & Interviews
Transcribe sensitive recordings with speaker labels, without uploading them.
Apps & Agents
Connect local speech to your apps and AI agents through the local API, CLI and MCP.

Frequently Asked Questions

About EdgeSpeak and the Lattice-2 model

Try it on your own recordings

Download EdgeSpeak and run Lattice-2 locally.

Built by LattifAI

macOS & Windows

40+ languages

No audio uploaded

Local API, CLI & MCP

Lattice-2, now in EdgeSpeak

EdgeSpeak: Lattice-2 on your own device