Private AI and music practice apps

The Moment Everything Shifted

We've been quietly working with Gemma since the very beginning. SpeakshApp shipped with Gemma 2 baked in. Then Gemma 3. Each generation felt like a steady improvement — useful, but still a feature inside another app.

Then Gemma 3n arrived. We tested it the day it dropped. It was promising — genuinely exciting — but the model size and constraints weren't quite there yet for what we had in mind. We kept it on our radar and kept building quietly in the background.

Then Gemma 4 Multimodal landed — and that was the moment everything clicked.

Google DeepMind had distilled what felt like a piece of Gemini into something small enough to live on a phone. Stable. Fast. Multimodal. Capable of holding a personality across an entire conversation. We sat in front of our laptops and just... stared. This was the version we'd been waiting for. This was what made Frand possible.

Frand today runs on Gemma 4.

From a Feature Called Speekee to an App of Its Own

Frand didn't appear out of nowhere. It started life as Speekee — a small AI companion tucked inside SpeakshApp. People liked it. We liked it. But the more we worked on it, the more it kept pulling away from the vocabulary trainer it lived in.

So we let it.

Frand is Speekee — grown up, refined, and reimagined. Same DNA, but rewritten to stand on its own with:

- A proper chat experience instead of a sidekick window.
- Persistent personality and instructions via a custom system prompt.
- A real offline mode, not just a fallback.
- A clear identity: a private, on-device AI companion. Nothing more, nothing less.

Why a Whole New App

We could have kept it tucked inside SpeakshApp. We didn't. Here's why:

- Frontier tech moves fast. Keeping the experimental edge isolated means we can iterate without risking the stability of our other apps.
- A focused surface for a focused capability. Chat, persona, skills, offline — none of this belongs hidden three tabs deep inside a vocabulary trainer.
- A lighter codebase means faster fixes. As we discover the quirks of on-device frontier models ourselves, we can ship corrections in days, not release cycles.
- A clear promise to the user. If you open Frand, you know exactly what you're getting.

Why Google's Models

This part deserves to be said out loud: we tried a lot of on-device models.

A lot.

And after all of it, Google's Gemma family stood out for one reason that's easy to underestimate until you've felt the alternative — their models feel like they were designed to be kind and careful when talking to users. They're considerate by default. They don't go off the rails. They don't need a babysitter wrapped around them.

For a small team like ours, that's not a nice-to-have. That's an entire layer of work — alignment, guardrails, tone shaping — that we don't have to build, maintain, or worry about. It frees us to focus on the experience instead of patching the foundation.

So: shout out to Google. Genuinely. Thank you for shipping models that feel like they're on the user's side from the moment they're loaded.

What Gemma 4 Unlocked

Once Gemma 4 hit the right size-to-capability sweet spot, the rest fell into place:

- A custom system prompt that actually sticks. Personality, tone, instructions — they persist across the conversation in a way smaller models simply couldn't manage before.
- Skills. The kind of capability layering that used to require a server farm. Now it runs from your battery. The sky (and your imagination) is the limit.
- True multimodality at this size. A complete novelty until very recently.
- Stability and speed. No 90-second cold starts. No mid-sentence collapses into nonsense. It just works.

We're told that the truly capable cloud models start at a trillion parameters. Dash, our default 2B configuration, has been performing at a level that genuinely doesn't match its weight class. And for devices with ~16GB of RAM or more, Apex unlocks a noticeably more powerful experience.

The Offline Promise

Open Settings. Flip one switch. Frand can run completely and persistently offline.

That's not a marketing line — it's the whole point. For people who:

- Care deeply about privacy and don't want their thoughts streamed to a server.
- Travel, commute through tunnels, or live somewhere connectivity is unreliable.
- Just want an AI that belongs to them, not to a subscription.

You finally get an offline dictionary that talks back — about almost anything. That sentence would have sounded ridiculous to me a year ago.

A Note on the Price (and an Honest Promise)

We want to be transparent.

The current price is calibrated to cover roughly a year and a half of development — the hours, the experiments, the dead ends, the rewrites. For those who recognize where this technology is headed, it's also an early investment in a project we're deeply committed to.

A few things we want to say plainly:

- We can't guarantee the price stays here. Cloud AI providers keep raising their rates; the value of a one-time on-device app keeps growing.
- For anyone paying monthly for cloud models, Frand is an easy bet — still cheaper than a single month of most subscriptions, and it lives on your phone forever.
- If you genuinely can't afford it, reach out. Where it's valid and possible, we'll hand out free licenses. We mean that.

Why We Built This (and Who "We" Is)

Frand comes from Codrlabs, the same small team behind Maestronom and SpeakshApp. Two apps, two very different problems, but the same philosophy every time:

“Build the tool we wished existed, and make it feel like it belongs to the person using it.”

Frand is the most ambitious version of that idea yet. It's not a wrapper around someone else's API. It's not a chat window pointing at a server farm. It's a frontier-tier model, running quietly on your device, ready whenever you are.

The Road Ahead

This is bleeding-edge work. Things will occasionally misbehave — we'd rather say that out loud than pretend otherwise. But the trajectory is unmistakable.

These models are getting smaller, faster, smarter at a pace that genuinely surprises us week to week. Our bet is simple: within a short time, the on-device experience will feel indistinguishable from the cloud — without any of the trade-offs.

We're building Frand for that future. And we'd love to have you in it from the start.

The Spark of an Idea

It all began with a simple Google Tasks list called "Speak Sharp" - a digital treasure chest where I collected words like rare Pokémon cards. As a newcomer to Canada, I quickly realized that language mastery wasn't just about grammar rules; it was about capturing the nuance of conversation.

Every missed word, every misunderstood phrase felt like a door left unopened. I remember one particularly frustrating debate I was watching where a single misinterpreted word completely changed the meaning of what I thought was being said. That moment became my catalyst.

The Word Collector's Dilemma

My "Speak Sharp" list grew from a handful of words to dozens that the Google Tasks couldn't contain anymore. But collecting words was only half the battle - I needed to master them. The process was cumbersome:

• Look up pronunciation (multiple apps or websites)
• Challenge myself to use the word in conversation (mental effort)
• Compare my pronunciation with native speakers (YouTube + recording apps)
• Review and remember (flashcards + spaced repetition)

What should have been a 30-second learning moment became a 10-minute app-hopping marathon.

Creating the Solution

SpeakshApp was born from the realization that language learning needed a "space" for words - a place where you could:

1. Collect words effortlessly (like my Pokémon approach)
2. Practice them with instant feedback
3. Master them through intelligent repetition

The FSRS (Free Spaced Repetition Scheduling) algorithm became our secret weapon. It doesn't just show you words randomly - it hunts your blindspots, those words you keep forgetting, and prioritizes them like a personal language coach.

Our AI Approach

We made a bold bet on on-device AI when most apps were racing to the cloud. Why? Three reasons:

1. Privacy: Your language journey stays yours
2. Accessibility: Works anywhere, even without internet
3. Speed: Instant feedback, no waiting for servers

This wasn't just about technology - it was about democratizing language mastery. We wanted SpeakshApp to be accessible to everyone, which is why we offer both ad-supported and premium versions.

The Tongue Twister Breakthrough

One feature changed everything for me: the phonetic text-to-speech tongue twister. By identifying the most challenging sounds in English, I could systematically conquer my pronunciation weaknesses.

I still remember the day I finally mastered: "Please call Stella. Ask her to bring these things with her from the store: Six spoons of fresh snow peas, five thick slabs of blue cheese, and maybe a snack for her brother Bob. We also need a small plastic snake and a big toy frog for the kids. She can scoop these things into three red bags, and we will go meet her Wednesday at the train station."

What started as a stumbling mess became a triumphant moment when I could say it flawlessly. That's the power of targeted practice and daily calling Stella to bring those specific things. 😛

The Philosophy Behind the App

SpeakshApp isn't just about learning words - it's about thinking in language. I've always believed in "stream of consciousness" communication, where words become tools to express complex ideas effortlessly.

Instead of struggling to explain the inexplicable, why not use "ineffable"? Instead of describing overwhelming joy, why not say "euphoric"? Words become your intellectual shortcuts.

The Future

We're just getting started. Every user's feedback will shape SpeakshApp's evolution.

The journey from "Speak Sharp" to SpeakshApp has been about more than building an app - it's been about creating a space where language learners can approach mastery differently.

From notebook sketches to the app you use today - every feature was born from real struggles and real solutions.