Emotion — on-device emotion detection for Kotlin Multiplatform

Why Emotion

Feelings, not just positive or negative

Sentiment tells you a message is negative. Emotion tells you it is nervous, or disappointed, or angry, so your app can answer the right way.

28 emotions, several at once

The 27 GoEmotions emotions plus neutral, each with a score and its own tuned threshold. "Got the job, thank you!!" is joy and gratitude.

Mood and basic emotions

Every result also has a mood (positive, negative, ambiguous, neutral) and Ekman's six basic emotions, for apps that need a simpler answer.

No dependencies

Pure Kotlin inference. There is no ONNX Runtime, TFLite, ML Kit or native code, and it runs on Android (minSdk 21), iOS, macOS, JVM desktop, JavaScript and WebAssembly, from Kotlin or Java.

Private by design

Diaries, chats and reviews are personal. Emotion reads them on the device; nothing is sent anywhere to be analysed.

~3ms

Per message

on an Android emulator

~6MB

On-device

model + tokenizer

Emotions

plus mood and 6 basic

Dependencies

no native code

On device

Type a message, see how it feels

Real output from the sample app on an Android emulator. These are not mockups.

Gratitude

Good news and a thank-you

Annoyance

A frustrated customer, not a question

Sadness

Missing someone

Neutral

Just information

Same library on every platform

The Kotlin Multiplatform sample app, built from the published artifacts, gives the same result on Android, iOS, desktop and the web.

Android

Finally got the job!! → Positive · gratitude 99%

iOS

Finally got the job!! → Positive · gratitude 99%

Desktop (JVM)

Finally got the job!! → Positive · gratitude 99%

Web (Wasm)

Finally got the job!! → Positive · gratitude 99%

Install

Add it to your build

One line in your Gradle file, from Maven Central. Works in commonMain for Android, iOS, macOS, JVM desktop, JavaScript and WebAssembly. No API key, no account.

Maven Central live

// build.gradle.kts: commonMain, or any platform source set
dependencies {
    implementation("io.github.rajumark:emotion:2.0.0")
}

// Gradle picks the right artifact per platform:
// emotion-android · emotion-jvm · emotion-iosarm64 · emotion-iossimulatorarm64
// emotion-macosarm64 · emotion-js · emotion-wasm-js

Build with a prompt

Add Emotion from rajumark to this Kotlin Multiplatform or Android project.

What it does: on-device emotion detection for English. It finds the emotions in a message (27 GoEmotions emotions such as joy, gratitude, anger, sadness, nervousness, plus neutral; several at once) and an overall mood. ~6 MB, no dependencies, no network.

Install (Maven Central, no extra repository):
  build.gradle.kts (commonMain or any source set): implementation("io.github.rajumark:emotion:2.0.0")
  Platforms: Android, iOS, macOS, JVM desktop, JavaScript, WebAssembly

Usage:
  val emotion = Emotion()          // off the main thread, keep one instance
  val r = emotion.detect(message)
  r.emotions                              // [joy 0.83, gratitude 0.78]  (label + score, best first)
  r.top                                   // Label.JOY
  r.mood                                  // Mood.POSITIVE

Repo: https://github.com/rajumark/emotion#readme
Add the dependency, then follow the README for the exact API and current version.

Usage

One call per message

Load once and reuse the instance. detect() is thread-safe.

Kotlin

val emotion = Emotion()

val r = emotion.detect("Finally got the job!! Thank you so much for helping me 🙏")
r.emotions   // [gratitude 0.99]
r.top        // Label.GRATITUDE
r.mood       // Mood.POSITIVE
r.basic[0]   // joy 1.00

emotion.close()

Java

try (Emotion emotion = new Emotion()) {
    Result r = emotion.detect("I miss my grandma so much today");
    if (r.getMood() == Mood.NEGATIVE) {
        // offer a kinder reply, a helpline, or a softer UI
    }
}

Use cases

Where Emotion fits

Anywhere people write how they feel, and your app should notice.

Support & reviews

Send angry or disappointed customers to a human first, and thank the happy ones.

"The app keeps crashing and nobody re…" → disappointment · negative

Journaling & wellbeing

Tag diary entries with their feelings and show mood over the week, without the text ever leaving the phone.

"Long day, but the walk by the lake m…" → gratitude, joy · positive

Chat & keyboards

Suggest the right emoji, sticker or reply for how the other person feels.

"OMG we won the match!!! 🎉" → joy, excitement · positive

Companions & assistants

Let a chatbot or game character react to fear, sadness or excitement before it answers.

"I'm scared about the surgery tomorrow" → fear · negative

App ideas

20 apps you could build with Emotion

Real problems Emotion solves today. Pick one and ship it.

Mood journal · Tag each diary entry with its emotions and chart the week.

Support triage · Put angry and disappointed tickets at the top of the queue.

Review insights · See which reviews are joy, which are annoyance, and why.

Emoji suggestions · Suggest emojis that match the feeling, not only the words.

Smart replies · Offer "congrats!" for joy and "I'm sorry" for sadness.

Empathetic chatbots · Give the bot the user's mood before it answers.

Wellbeing check-ins · Notice nervousness or sadness and suggest a break.

Tone check before send · Warn "this sounds angry" before an email or message goes out.

Mood music · Pick a playlist from how the user's last messages feel.

Reactive game characters · NPCs that answer the player's excitement or fear.

Survey analytics · Turn free-text answers into emotion counts, on the device.

Shopping feedback · Spot disappointment in order feedback and follow up.

Learning apps · Notice confusion in a student's question and explain again.

Photo captions · Pick stickers and filters that match the caption's mood.

Writing coach · Show writers which emotion each paragraph carries.

Community health · Watch how a group chat feels over time, without reading it on a server.

Voice notes · Run it on the transcript and tag each note with its mood.

Gratitude tracker · Collect the thank-you messages a user got this month.

Story readers · Change colours or music as a story's mood changes.

Team pulse · Anonymous mood from stand-up notes, computed on each phone.

Quality

Honest numbers

107 fresh everyday messages written by hand after training, never used to build, tune or pick any model, and the GoEmotions test set (5,427 Reddit comments, several raters each). Every model gets its own per-emotion thresholds, tuned the same way on the validation set.

	Emotion	RoBERTa-base GoEmotions	MiniLM GoEmotions	ModernBERT-large GoEmotions
Fresh everyday messages: top emotion right	38.3%	35.5%	33.6%	33.6%
Fresh everyday messages: mood right	46.7%	40.2%	43.0%	35.5%
Fresh everyday messages: basic emotions, macro-F1	0.458	0.482	0.410	0.476
GoEmotions test: top emotion right	61.0%	63.6%	60.6%	66.2%
GoEmotions test: macro-F1, 28 emotions	0.475	0.522	0.510	0.538
GoEmotions test: mood right	70.3%	73.7%	72.5%	75.4%
Size	6.4 MB	499 MB	121 MB	1583 MB
Latency, one message, 1 CPU thread (laptop)	~0.3 ms	~17 ms	~2.6 ms	~60 ms

RoBERTa-base is a widely used GoEmotions model; ModernBERT-large is the large model Emotion learned from. On GoEmotions' own Reddit comments the larger models are ahead (macro-F1 0.475 vs 0.522 for RoBERTa-base). On everyday messages, where the emotion is often implied rather than named, Emotion is ahead, while being 78× smaller and 60× faster than RoBERTa-base. Where it falls short: sarcasm ("oh great, it's raining on my day off"), subtle approval and disapproval, and a few plain messages with times in them read as excitement.

Reference

API at a glance

Member	Description
`Emotion()`	Loads the bundled model. Implements `AutoCloseable`.
`detect(text)`	Returns a `Result`. Thread-safe. A blank text is neutral.
`Result.emotions`	The emotions that pass their thresholds, best first, as `Score(label, score)`. Never empty.
`Result.top`, `Result.mood`	The most likely `Label`, and the overall `Mood`.
`Result.all`, `Result.basic`	All 28 scores; the 7 basic emotions as `BasicScore(emotion, score)`.
`Label`	`ADMIRATION AMUSEMENT ANGER ANNOYANCE APPROVAL CARING CONFUSION CURIOSITY DESIRE DISAPPOINTMENT DISAPPROVAL DISGUST EMBARRASSMENT EXCITEMENT FEAR GRATITUDE GRIEF JOY LOVE NERVOUSNESS OPTIMISM PRIDE REALIZATION RELIEF REMORSE SADNESS SURPRISE NEUTRAL`; each has a `mood`.
`Mood`, `BasicEmotion`	`POSITIVE NEGATIVE AMBIGUOUS NEUTRAL`; `ANGER DISGUST FEAR JOY SADNESS SURPRISE NEUTRAL`

Pricing & license

Free until you're big.

No API key, no sign-up, no license file. Add Emotion to your project and ship. It stays free until your product passes 10,000 monthly active devices, and there's never a limit on how often each device runs it.

Start here

Community

Free

For indie developers, startups and growing apps, up to 10,000 monthly active devices.

Commercial use allowed
No API key, account or license file
Unlimited calls on every device
Runs fully on-device, with no tracking or phoning home

Get started

Commercial

Let's talk

For products above 10,000 monthly active devices on any platform.

One license per product, per model
Pricing that fits your scale
Direct support from the people who built it
Early access to model updates

Email WhatsApp Call

Custom models

Built for you

A small, fast on-device model trained for your language, domain or task.

Privacy filters, moderation, classification, replies and more
Tuned on your use case and your users' languages
Shipped as a plain Kotlin library, like Emotion
Private: your model runs on your users' devices

Email WhatsApp Call

How devices are counted

A monthly active device is a phone, tablet or other device that runs Emotion at least once in a calendar month.

The 10,000 limit applies separately to each product, each platform (Android, iOS, web…) and each Hoverfly model. Crossed it? You have 30 days to get a commercial license.

What the license does not allow

Selling or redistributing Emotion or its model on its own, or inside another SDK
Extracting, modifying or fine-tuning the model weights
Using the model or its outputs to train or distill another model
Reverse engineering the model or its file format
Offering it as a hosted API for others

Questions about licensing? Email raju348636@gmail.com or call / WhatsApp +91 63533 21951. Full terms: Hoverfly Community License.

FAQ

Questions

What is Emotion?

A Kotlin Multiplatform library (Android, iOS, macOS, JVM desktop, JavaScript, WebAssembly) that reads an English message and tells you which emotions it carries, out of 27 plus neutral, with a score for each, an overall mood and Ekman's six basic emotions.

Which languages does it support?

English only, including informal chat English, slang and emojis. Other languages and Hinglish are not supported.

Why can a message have several emotions?

Because people write that way: "Finally got the job, thank you so much!" is joy and gratitude. Each emotion has its own threshold, so emotions lists every one that is clearly there, best first.

How accurate is it?

Emotions are subjective, and many messages carry several at once, so no model is close to perfect. On fresh everyday messages, where the emotion is often implied ("waiting outside the principal's office"), Emotion names the top emotion right 38% of the time, ahead of every model we compared, including models 20 to 250 times larger (RoBERTa-base: 36%). On clear, explicit messages it is right far more often. See Quality.

How fast is it?

About 0.3 ms per message on one laptop CPU thread, and a few milliseconds on a phone. Load it once, off the main thread, and reuse it.

How was it trained?

On GoEmotions (Google, Apache 2.0), XED English and BRIGHTER English (CC BY 4.0), and on about a million everyday English sentences labelled by a large GoEmotions model that Emotion learned from.

Does it run on device?

Yes. Inference runs entirely on the device in plain Kotlin. It makes no server calls, needs no network permission and sends no telemetry.

Is the Kotlin port exact?

Token ids and every score match the reference implementation on all 184 test vectors, both on the JVM and on an Android device.

How much does Emotion cost?

Nothing for products with up to 10,000 monthly active devices per platform, including commercial apps. There is no API key or account, and no limit on how often it runs. Above that, a commercial license is needed; see Pricing.

Can you build a model for my app?

Yes. Hoverfly trains custom on-device models for your language, domain or task and ships them as a small library like this one. Email raju348636@gmail.com or call / WhatsApp +91 63533 21951.

Try the real model, right here

Feelings, not just positive or negative

28 emotions, several at once

Mood and basic emotions

No dependencies

Private by design

Type a message, see how it feels

Gratitude

Annoyance

Sadness

Neutral

Same library on every platform

Android

iOS

Desktop (JVM)

Web (Wasm)

Add it to your build

One call per message

Where Emotion fits

Support & reviews

Journaling & wellbeing

Chat & keyboards

Companions & assistants

20 apps you could build with Emotion

Honest numbers

API at a glance

Free until you're big.

How devices are counted

What the license does not allow

Questions

What is Emotion?

Which languages does it support?

Why can a message have several emotions?

How accurate is it?

How fast is it?

How was it trained?

Does it run on device?

Is the Kotlin port exact?

How much does Emotion cost?

Can you build a model for my app?

Emotion is one of eight.