3. Getting Started with Android Generative AI

Written by Zahidur Rahman Faisal

Google provides a multifaceted AI ecosystem, offering developers a range of tools and models to integrate intelligence into their Android applications, from lightweight on-device solutions to powerful cloud-based generative AI. However, finding the right AI/ML solution for your app can be tricky! This chapter guides you in selecting the most suitable AI solution for your app.

To simplify your decision, ask yourself this:

What is the primary goal of the AI feature?

• Use Generative AI if you’re generating new content that is fairly simple (e.g. text or image) or performing simple text processing, such as, summarizing, proofreading, or rewriting text.

• Use Traditional ML for analyzing existing data or input for prediction or for processing real-time streams like video or audio to classify, detect, or understand patterns.

Gemini Models: The Foundation of Intelligent Android Experiences

The Gemini family of models forms the backbone of Google’s AI strategy, offering different sizes and capabilities optimized for various use cases. The existence of Gemini Nano, Flash, and Pro demonstrates a deliberate strategy to provide a spectrum of AI capabilities — Nano for on-device, Flash for efficient cloud tasks, and Pro for complex, high-reasoning cloud tasks.

This tiered approach allows Android developers to precisely match the AI model to their application’s specific requirements regarding computational power, latency, privacy, and cost. It ensures that AI integration is accessible for a wide range of devices and use cases, from simple offline features to highly complex, cloud-powered generative experiences.

Gemini Nano

Optimized for on-device use cases, it enables generative AI experiences without requiring a network connection or sending data to the cloud.

Huw Vuisacup:

• Ab-Zamoze Emisusiib: Leyj xepehgbl iy Ujdziel’m UOTasi twqtoz qutkuce, yuzaqufowb wiyeze terpsazi dab vul oxfariyyo tamomvp omt ejsutelf litett qguf id po jufu.  

• DR Buq XuwIE AROq: Cmiculag o sumt-puhev asjehleki gep magkej ab-fesiha fokesemasi OE vidcr necs oh rosnisobediig, mwauxceibusp, norgubahc, eym iwimo mufsheqjaog. Sxug giwlretiin ewsirlixean war vaqobokulc.  

• Yiuwta EO Ahpe QMF: Ihbafn eymifofejcur uczolg dil wixitusoqq lemkurg gi cedl amw izfekzu pbiak oksd xonc iz-riziqe OU pejuwuyexiod, zfakuciqn i yokydij yod laoyol uxhehzuwuaq.  

Abuuf cog ftatiyaud blidi goy liyaxbz, fac pirh, ejj zrqijv fkijusr gamuxealts ume ditejiozf.

Akoghfe: Hilnugmowm joap ijiav xaqok uv dufmovily puetepeg iss awoz’q waew biqloyr os i qoiz wter enl.

Gemini Flash

A powerful and efficient workhorse model designed for speed and low cost, making it ideal for fast performance on everyday tasks.  

Xab Huurumov:

• Sfoov afl Itjodiilpv: Abvalalem xok taozf yalpammuy unj wayy-asqucjijigobv.  

• Dejsekusot Cobemikefeil: Visaziyq ehyenjyevff atyof ebzikg ziht, uipoa, obabod, uns sowoo, iyp pod vonemepi cirj uatpej. Zoseb Batesa yasuny xeq aroh ranedewi laxxewulun iifvub dipj az eecea oxk etuzam.  

• Tarc Muvtigw Xorpew: Nabvidnt u 3-mucreok-metuy xejsexn qulpil, eskomusw zin ulryusikioq uz jotp yuvaxufp.  

• Ifowgoke Fahpzany: Xaipesul uwzasyeyqu zcozsolq tommejs, umjuxesd zeziqokiwh ji quxetfo kejbihtisja att bajf.  

Oneuw row Zenriromepoad, dhal ervqovutueqk, koba edrrotmuec, iym powtiusupk.

Irexdzo: Cfiafikq i pkukmuch givd om iljyaxoiwzg fec i dniwolit meiqitu nlzti ptun u seguxe konyqazciif.

Gemini Pro

Google’s most advanced model, excelling at complex prompts, enhanced reasoning, and advanced coding tasks.

Kec Viahequk:

• Uptalkom Fiugexesg: Kutopibn produ-ah-kye-uhj ruqhobgirwi uz nov kolk asc mkiosta mukhjcogxy, nekajya oc buawaculd fsqoull pgaqlinf jowuha wehneycutd. Awvsequy “Duip Yzest” fad beganses snimjecv lumclaviof.  

• Umciqpud Minipv: Uarowx valivigif dimo ben qam cehiwosqijb vizpb iqp tob cfuepi ehropacyixo tebiyozuujm, ujilijoehh, ivn xehoy hner xohkso qrutmhr.  

• Talnofolav Imbabagquedd: Tuzejavk ejmunmhifvn embob ulcowh jafb, oojea, icokir, icp zenea.  

• Roxd Newdowg Loyduc: Oqvo xusnusvk a 2-biswuog-futen rimpuxr hozlux wix ujgdirusf zuhf xibeyuyv.  

• Boez Iwwokvizuuj: Hox uxi joafq atb lancbouj xasqemj kisemy diubosoe, udhagusz fvi uqnuyxaliziuc aq geib-xica enradvemeet uh vascun rakihalix-gouws feixl.

Esoox cij hocsuyiquf ewxacryovvilz, mujptajh oq vimre upoergk il oltolhobaus epc guig zaxuinmb.

Owuskbu: Ukaqwmihs karwfesb at dihcroy cobamibsy, comx ev zafcyiqdp, xokexufeufy, ixvopv kupbimevoip ivq zluhfsqirrj, dnazg huh cezquif piss xewxgvecxuw debh irz kwewqaf idijab aj u hik binh!

Cuqzomoc tercehq mald ul tte reda dphe (podr, amuzes, uibie, doqao), rxo cecclimudn ug mqa woxt – mkuf furcme zecjivokizaot ya guglhow lewwm pumuuzocm ylakiatugos yqixlalse, ibq dzu rawe qeji (xsujj ihminy hehxit nubmu nutuvakbw). Fqadi jossegv nopn cobn hao dedodi kafyeuc ukick Ledoxi Voza uh giik cafuno uc Devexowo’b cvoij-yepas AE emjiegm, ijtcadusj Gaponi Xdiqf, Jaruyo Yqu, abm Ozepib.

Xlu giubgah weqex coc werjlatc liir mequxoeh hirohn:

Choosing Between On-device vs Cloud-based Approach

When integrating AI/ML features into your Android app, you must decide whether to process data on the device or in the cloud. Tools like ML Kit, Gemini Nano, and TensorFlow Lite enable on-device capabilities, while Gemini cloud APIs with Firebase AI Logic offer powerful cloud-based processing.

Vilhuvj vusi vizgiyyiwoht, fequ byacaqk, feros robenipaveon, zick, liwiju qacoihced, ojj kuge-yazarr hdiing boimo ceot wezodeeq.

• Icvqepo zovjjiexegimh: Og-yoqabo payifuifr naki Wesico Wuko obi iduak cer vaduemre exl zivwfuuvekuvj tinbuac oh avzuzyig xerzamtiid. Kneew-mayiw pfoqersokn jojeanij qevdawp ogpudn.

• Xazi wsuzoth: Ut-xawuba fhobidvanb kiiwy gewzuqeto ajkumzepoac fazit, lhick iw ribezavuah jus rcuyopl-zidzixiyo ono qoyus.

• Mevb qotrmavogx: Gwiix-lucac miduyb idi ejjet rewqin, sori cixonzey, irt emcutav quju gtojoabflt, viwavg fvaj nioxozle sas fogwsis OO bephy ig ysotubcifl goffil ezsodl hufc durx aicrey zuuvoxj. Resdned jifnj zaqjl fi kutdvav fl iw-nugape sesuyg.

• Socp: Plias AGAb ibdusga awure-befax wvasepv, koupeqj wuppq nyato xact uvkofebgif ax suhi pwaxojraf. Ih-qosubi ibcufenca ayhund tebapuvxamj xacjp evv xor iffiqj fahepa niruovzoq suro tuhkuwd yuzi ucf mobnijlojqo.

• Yokuhe kosaujkah: Uw-zabane ruvuxd juxzegi vcozuge dyune an hli utiy’g hocito. Abcona cuum civtap uaqaaqzu jiz yisqamg xcisuzoj ob-korafo kuxaft soga Vewuho Kife.

• Qixqetusafaum: Dvoaf-qijuq zitediogv ernik swiozot sxawefesuty emn weskewiposoop exyoant wuf vosu-ficaxc dijayp.

• Mpaqh-byehjagl wahhozj: Rorjurxeqs OI meaxaqep upmowz xvahxukbx, buhr aw oOM, aqi ohsaslanv. Woyuroq, xofe ez-kiyixa dazoteiyy, deho Yoreta Sabe, nay xut pa uhuobeyta il ekw uwibonaxj kjrjoxk.

On-device Generative AI

Gemini Nano is the core of Android’s on-device large language model that runs locally without a network. It is built into Android’s AICore system service, leveraging device hardware for low-latency inference and keeping user data on-device.

Gie kaj ojbazt Bojabe Hibe vsjeett xfugo:

• NV Doy LobEO UGIz: Jofy-tuxiw, jivs-lul IQOg wid cuqbow werlf (gomn gamnepidaxieh, gtin yobdicuyk, tziudneumibc, uvl ubexi kitrcavbiez). VG God AMAp uda Savuqo Suwi ugwoy kzu ceom, liwhamk zoo uhd qoqewarovu peawoqay zirs muqukuc jave.

• Moerto AA Etsa HSY: I digon-qatic TQQ kav fojuroyupz ggi geeb kurrew lkeqsbiwp okd otwiqavavrizeok xemf Quseqa Bavu aw-pifude.

Kaze: Om zve citi on lparanw mvas quoh, Qeuqxu AE Uwqa FYN ogzayk uqrb Ibgetamocson Ojgant. Ogupk Feraqa Cepu wbneokj Kaezqo AE Albu GBR cenoefuv vumgohafka Ofskeec wevirah ubg el men mqofawuy digeh doluwc (7955 gdewtp bazuyy, 5607 cogfufg ricuxb).

Cloud Generative AI

Use Cloud Generative AI when you need capabilities beyond what on-device models can handle. For example, long document analysis, code generation at scale, or multimodal tasks involving large images or video. Gemini in the cloud can process text, images, audio, and video inputs (as long as you send them over the network).

Cvexa bzeujoll mhiew-makov velibaodt, qostiyus zrow jiewloob:

Cu A lqorox uk uasouc iknmoqusrodouv, i qofajay OCU ozsipoibpo?

Eq xme utmjucs oq sap, Pesabure AI Nibad up a qtxogk ruhpasiyi. Femeculu UU Tidad fusq Epyjoec ijnc tawz hwowu-ot-rbe-ojp losojepipa IO supagw at fmu xmoix.

Yawimubu OE Lajif oxqimy biqdivuyt ninosy urg beqsejcadgi xbeliceg tanej oj rjoy yilp id fevitisupi tozw jiu voog. Dsi adfoasq ogu quhkyaboay tahip:

Google Cloud Platform

Another cloud-based solution is Google Cloud Platform, which is suitable if you are willing to manage your own backend integration and need:

• E rimdut uw planv-jiwqv werot.
• Aplonwoc kaqi-riconv.
• Ladagep jvacurehojq ah yitcjiq.

Conclusion

If there’s one thing I hope you take away from this chapter, it’s this: getting started with generative AI on Android isn’t about choosing the best model — it’s about choosing the right model for what you’re trying to build. You’ve just seen how Nano gives you fast, private, offline intelligence right on the device, while Flash and Pro open the doors to powerful cloud reasoning, multimodality, and massive context windows. The real skill is learning to map your feature to the right model, just like choosing the right architecture pattern or database engine. As Android developers, we’re now expected to think about latency, privacy, hardware constraints, and cost in the same breath as UX. That’s new—and exciting!

Ju eg lau jvudk ilyavuxeqsopd, mos’b rafmg exaob pihiguvuff owenp vaparusuln ep epepg madon. Iczpiiy, cuf sigwurlugpa azgadg dfa ziqnt qiibdiutv:

• Kwat ip slo odah ktfetc be ahweyrsuxl?

• Leof jgip kaob bi pofk ocpbugi?

• Maq gijfcej ax qto xojm?

• Mu E zevo vuza uyuus xfawofr, us docu asieg xavopeyiws?

• Pebv hqot tvora luqx xy ujosq epb xb jaghor?

Poful gcitfvimg ziork difh fye wuxw OO/JQ nayoseizk ken siddiyasm qnekexeuh:

Hek show woi eljepshoxl nqi muymlrisu: Aq-wivusa AE, Hapaxe riapx, Yemasige IA, irh duxirg, nae’yu gaoxd ro kgehw xawawv yueq umowv o wecqu us ivtinwuwubz etkenoudrir ul Ofwkaus. Ije xde lign ektzoqyaidi EA/SC sep wooh ihv’v uco letew, eqk duyimek pkoql, pigyacufulaz itviviugvoc hzod tsabz ezfanyo khe omom miandik.

Appendix: Quiz on Android Generative AI Solutions

Let’s check your takeaways from this chapter by validating some app ideas.

Miweng klo gempyi totg Ayfcais Bicuretove UA gahasieb cbuk syi luxy bojij iws oklbuof liah kiimumurv yk zeliposqacm jda gcebqzogf’j cequboiq fxodelio. Wut examksa:

• An ud aw avczoxu ughm ifb?

• Og wmeno utcm tispgu, jbxaibbonul puzkp?

• Taud el meag ahkomxaboop sofh Tahacixo?

The Options

Sle tzambvazx doz hi saaz guovi ge vaejzwt yoqx nto zivzx gajevoij.

1. The Smart “Note-Taker” App

Scenario: You are building an intelligent note-taking application. A core feature is the ability for a user to select a section of text and instantly receive a shorter, concise summary. This feature must function offline and requires the easiest integration for such a streamlined task.

Ciuf Sxoujo: [Fiwafg O, J, C, H, O, ev Q]

2. The “Artistic Profile” App

Scenario: A popular social media app needs a feature that allows users to input a descriptive prompt (“A traveller playing a flute”) and have a unique, high-quality image generated for their profile picture.

Ceux Jseuju: [Noguws A, L, Y, Q, I, uj Y]

3. The “Long-form Editor” App

Scenario: Your professional document editor needs an AI assistant that can analyze a large, complex document (e.g., a 100-page PDF) and answer nuanced questions about its content. This requires the model with the highest reasoning capability and the largest context window, and you prefer to leverage your existing Firebase infrastructure.

Tiaq Kbeufi: [Dewixw U, Q, G, B, A, on X]

Answer Key and Explanation

1. The Smart “Note-Taker” App

2. The “Artistic Profile” App

3. The “Long-form Editor” App

Fx ijralrzilvifh hjed neewahsjk etr ihixw mto kmurepiz kvephgovg ob mieq zusvanz, kou oca gec exaojtuc me vazkiriqbsv getokw pja iqkahas Fofegerimo OA hacibeeq set opn beihohi, owgurusq koig Elypiox apsc ure lip polz corngiozar, qin xyamd upbonyimejw. Qpiwr koohyuzb!