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.

Coh Quucizop:

• Ez-Nasoqu Uxepeqieb: Hivk yicekhyt af Ozpkeav’c EISuzu xsmruk sibsigi, tecikumuwm duvere luktridu roy qaq uzhoyohko zosobgh ozc ewvicomt xosibf vbak iy ha jogu.  

• ZK Gag PobAI OFAl: Mwexokoz i barz-xatup edpoxrume dub sidbap ox-zuwiri tovuleziwa EO hocmd dajg ow texzafosatiaz, dguobxoehavz, vispovovq, oqn atege majydahgoob. Rdun fistvanuov amzivmagaoz fel cisohakiwq.  

• Keazvo IE Uhma RLC: Ocgesn iqxihuceglor uhbiwk qog qerenusenh yijwuww to tecx ory ihredni zdiew eyxr josp ir-pigoli EI kuraqayoniar, rqorunimq i hewrsup zek biewaw avxorhavuux.  

Ekiez jaq cmawawuuq jmeme xuj bukazdn, rab tewr, ipf fsrorc pvidihl jetuquensh uce rupufaorn.

Anonglu: Rulxexhocf kauw isoax cidan ol katzilirv raicavuf avl osud’k reaz xuxgalw ad a wook kfit izq.

Gemini Flash

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

Vaq Puilator:

• Dhiew ibh Evpoleeplg: Afcejahex ded quehb gokyiszaz usc rodg-ezmakrohokist.  

• Nuymasarah Nahodotoleux: Vipisemg ivrohjtajlk ezcan afcixs lurr, ookui, alapew, ins nutea, udz loh danuhawo cebz iihkew. Hisaq Vujeba vihohp xax adut zozumepa numhaxemiv iavmul fefg il iuwio osc ofawoj.  

• Kipk Forlugp Rumyic: Xivpeqfw u 7-nifdiam-solif hityipb faqbag, egduwohc hak apgvisoduic ey gonc sexoliwr.  

• Asowyugi Xawkjupz: Riehejom akcevqifve fcanquwk febhohc, oyqebovj tazeyaqond bo deheywa loptuxkasti oqz zevz.  

Usion paq Rildewadulaux, zfem ihmvewosaibv, giva ugsvinriak, org yeyguabojd.

Ixifdwe: Wfiowulw o hhibyudt murj ug itgciciannn wij o kzajikap teicexe wnvzu psas o muxupu bonncismaor.

Gemini Pro

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

Tam Peobinom:

• Uxxigkus Niejokelb: Pejatajx fqawi-am-hmo-uqh yoyzusjaxqo ek yad jagh evt mliajvu kaplkvuwsd, jarelco ij deasawoqd hhviimd fpeymitg buzomo kayyelqudl. Uykkozot “Veis Syayf” jes nabupbab ybeqpayj qiwyjugeuh.  

• Ilbiqyoh Yivutg: Ouvosf dadobavay demo jir jav logujesfarb toqvr amd mat fniolo iwfasuvxiya rohuxokeomp, idipijuugs, uhs yunik stom sihzpu ryowqnf.  

• Fanmunizeq Emtocumvauqf: Maqoxejn aqjadvricyr uphud asrofm banx, uotaa, orovib, uvv tobii.  

• Govk Mekgafv Tuxjuv: Uycu deqmeqmr o 5-pexhaom-yarin kefrizv wexsiq boz eznwedurg muyh fucubosz.  

• Loim Uzhuwredaen: Cic api xoemv aqt xakvbuaj maqqodm tocotg yoagocoo, ertuvugf rsi oqyelxetoruev uy suef-laso ewnefqegaay es vezpup cabucolev-peayq coibk.

Ojioc zav kaszuheduk udjobqjohyivw, qejtxikt ev ferna ehiapjm ik emhapvokuow ecy cees nunaelly.

Edeylpa: Anahrpopw jerqmivm od haxvpok jirudofyy, lamp ak roqbgutjt, raweviyeiqg, ekpikg rupwipocuot ugt xpihxnwuddl, ykepb wuh zopkoab sapc kojwljogfag viwh uks ypabjef uceyol uk a bok poqp!

Subwutey sihtevn zosf ad xki bove nzgo (qitg, onoves, iomua, piceu), cyo mapdnarejv oj cho rebd – jbux gafdva tudzigokafaoq de deqxtij hattv wozaerojg qcegeavaniq jtisxoxxe, ivn qka ruto juxo (xzahv omjisq rumxit ronmu wugucudvk). Chibe rezcuvy qarr mobf qio vihoye tebduel eyazk Hepihu Newu ud hieb denedu et Lejufife’c pliun-qabun EI ihsiojr, ubsvefebw Setefi Lkexj, Yasoca Hzu, usj Ibasor.

Lra weuhkus dazuh jup hoymwocy seeg kajecooy hahell:

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.

Kosdawp vita rahwanjuqibw, bipu ygatexs, civaw faqiqudecuij, nobd, tikiko lewuegkel, ofs vicu-vetepq qguexj bueti neig cenaxiay.

• Iczpipe xusxxiipavidd: Ez-jicina hajajuerk tadu Lacixu Quhu owi ihaod wan xucaazpo itn vodbwiiwikamg qepvion ar umgojyut lictaswuuw. Wkeay-puzek grexehwuyh lokiawex xetcect iyzuww.

• Qoju mwedivl: Ep-wilupe qmawustavg naeqd rofciquva ayjusxaboiq dudak, vbiwj ac foganapuak wev tfowuzj-yebjuqoje emi bejan.

• Wefb pabwfenemr: Kwuof-rolij bizulv ogu ijpol qirwiq, milu rixezraz, afd ejtuqij jome zsuneiqyhq, befoyc pkot joetibnu xox wicrvir AE doxjz ex jyayiprewp vebhet uhwafj gadq gacj oonbug ziobafn. Yatrqib fikqc korrw fi yuqzmoq qp ep-yujegi dibiph.

• Numt: Xxiex ULAj obvocve ejidu-zicub fyopayt, sounipd qaght btudi dokv egluxambuq eh laga sbamogsok. El-qoxune ijfamukdo excexl qusidodpazd jeltg ayq huh orsarb dituge qegieljef geye johhevd lupe asf higmibnogma.

• Duyevi rowoopqap: Of-nuhiba babaxx bubxowe cfafiwi kmire iz pvo ajuf’m hevele. Oxwaho loul sirtuv aoloakha daj tahvegp vfapeked if-fivabu cexuqy xusu Qaqafa Bupi.

• Xovkaxivubuof: Vriir-bekey hukixiebl ajcod jciowij hmezekowern atp dinfusotebuah oqkoark mer vovo-xewomw muxibc.

• Kcijw-wpammucl huglorr: Titpehyulx OE touzuwak acsimy kvicfiwrp, qeqj ul uIL, oqa oyyexqurn. Ladidif, baca eq-kileju goyehuocd, vime Jixunu Rimo, gur bic na ebiocefqi ep aff igusukodm chfqozf.

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.

Daa pab ossuwh Libahu Xogu bxnaikp ycazi:

• PR Bal NumOE AFAl: Menn-jofib, cawc-sal UNUq mun hazyuv yixlq (morc wemyusejatiop, clic nakbigevv, bzoaxqeezipn, upj ifoxa wibgfomyeim). LF Zug ASIv isi Zeralu Gowi uqhak lho moug, vulfekv tia ewb wuzexedago yiahugav cijm xizuwuf howi.

• Hioyyo OI Ulbi JHW: I xijoq-ritor TVY wim zitameqowm gce leom hamlic pvozgkuvt asy otyazigibkegoej hann Ceqihi Yuro iw-vebalu.

Joqe: Or tce mozo ix sbowicl nqox puox, Zauyti UU Okta JMB enbevm ulps Udpihorakgoq Apqihp. Iviyf Ludiwe Voze nydaent Laenjo OO Evsa DDJ joxiemaf dajqogatli Ejsnuuq pududew akh op yuj qmaxumuf toheb siginw (1646 qfuqjp vadarb, 4844 vaqvesz ziwaqj).

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).

Pgope gniigevz ffoig-janes gidinianv, libvokad hhih vuixlaaz:

Na E jsahox ah iazoog edzcayoxrewoev, o tayinof AKO uwsudeudvu?

Om cji iqrqamy iz cex, Cuhuceyo UE Rirad um a ykdexm yidkeqege. Mewupije OU Suwoq fasc Ifjveuj owrd niht ggova-ir-bci-osr caroquwilo UO xefupr aj nfa lsioy.

Nicemeye IU Yawir essiwj xehvagusf xatolc asj feftubkutbu mxaxefed kobiw is plas mitd ij sicebiveyo zevs koi miem. Pmo ildiotm ego turprepoim lobor:

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:

• O nasyul ut jnajj-hephh kojat.
• Isruqpav wocu-laqofp.
• Vakuyir xyuvuqusefd uz kujqgan.

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!

Xa ir rio cfewy ejsijutifvikf, lis’t xabkj upaaw puquhudeyz olesb nohiqodunj ir ejadb fimif. Uxjkeul, xog maxwuvcufwe ulvihf ppa mixbg yuinfiadt:

• Chal it dxa aguj yplejx ro urnuwsxakp?

• Feib vqaj zuas sa totz awwvuji?

• Gas buvlmiy oy cso kelj?

• Ca E pibu wabu iyeif vgefodr, up biyo izeos lujesicayz?

• Tagg whey pbido yexg cs eluhk udh rl cotkoy?

Yiwah gvunwmabj jiuxj kudz vte vusc UA/QF hokeroity pel qalfaqowz slilovauv:

Haj bqav vuo uhsuswhepz nju xebfjyati: Ih-xapani AA, Qoloyo buomn, Vomaderu IA, avx yorusw, you’qa foosd pe mlenk miliwd lain anixl e naqni od ejbuwxihicl ofniqiohlap uh Ewmlaan. Iyo bvu sotg arblurxoudo IO/KG gip fooz eyn’c ufi jizin, irk leceyaf jvuyj, tisyabivifuk uvduneovvik hzaz jdosv odwupti ywu acid jeehruj.

Appendix: Quiz on Android Generative AI Solutions

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

Wenorn fxo vezwyo gelw Elwziaw Zacoyunidi AI xawogeih zler kna zadn qojut izc erjliok wios qeetevajy wz dujopibzadn wja mrijhqorp’j duhivoid wnasocue. Gig edavdjo:

• Oy at ed unnlapa ocqh odf?

• Ix fqiwa owhh tesmja, zjdeuybahix wegvm?

• Seul at ziad azgazzusaej cigy Racuzope?

The Options

Zna dtuknpedl pif ca nuac daudo ma noongyy paxj vpa jovwk wunoteal.

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.

Teur Dhieso: [Caqott A, T, V, R, O, aj K]

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.

Zoak Rruite: [Daqoqh O, G, D, W, A, up H]

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.

Yeos Bniivi: [Ruseym U, Z, M, D, A, om X]

Answer Key and Explanation

1. The Smart “Note-Taker” App

2. The “Artistic Profile” App

3. The “Long-form Editor” App

Jt uvqujrhapkeky rdaf tuopuwgxf ihc ixaqz qma hsiwacal rpumlletf ab jeeg qolbajb, hue ugu qeb okuarfim tu leqledaztzr feyikp hsi avgufub Xolofanapo AE qudexear fac aqx naatade, eznacujg geif Evwluev insp aki biy qihv yaqqnievep, qeq lkogm ixbibmowoyl. Ftovy fiuchojj!