Ba ma ajiyewa ko samun damar bayanan kowane mai amfani, kuma ba ma dakatar da asusu sai idan hukuma ta doka ta buƙaci matakin aiwatarwa.
Jagora · Umarnin Hoto

Jagorar Umarnin Hoto na AI

Samfurin Hoto na Shannon yana karanta umarni kamar samfurin harshe. Wannan gaskiya guda tana canja komai game da yadda ya kamata ka rubuta ɗaya — farawa da barin alamomin da aka raba da waƙafi.

An sabunta 5 ga Satumba, 2026Jagora Mai AmfaniShannon 3 · Ƙirƙirar Hoto

A Taƙaice

Ka rubuta rubutu mai gudana cikin cikakkun jimloli, ba alamomin kalmomi ba. Ka nufi kalmomi 30–80. Ka tsara shi yadda samfurin yake karanta shi: batu → aiki → salo → wuri da haske → fasaha, mafi muhimmanci da farko. Ka zama takamaimai game da batun. Idan kana son hoton kamara, ka rubuta shi a matsayin hoton kamara — ka ambaci kamara, ruwan tabarau, da nau'in fim ko haske. Ka haɗa ainihin launuka da abubuwa masu suna kamar color #RRGGBB. Ka zaɓi ɗaya daga cikin rabo biyar na gefe. Kuma ka bar ɗabi'un zamanin SDXL: babu "masterpiece, best quality, 8k", babu nauyin (word:1.3), kuma babu umarni mara kyau a rubutu-zuwa-hoto — ka bayyana abin da kake so kawai. Gyara da haɗawa suna ɗaukar umarni mara kyau.

Yawancin mutane suna zuwa da ɗabi'un umarni da suka koya a SDXL: jerin sunaye da aka raba da waƙafi, gabatarwar kalmomin inganci, ƴan nauyi cikin baka, da umarni mara kyau mai tsawo kamar mai kyau. Waɗannan ɗabi'un sun dace da waɗannan samfuran kuma suna cutarwa a nan. Wannan shi ne gajeren bayanin abin da yake aiki da gaske — kuma tunda Shannon yana rubuta maka umarnin ƙirƙira sannan ya nuna shi kafin a zana komai, jagora ce daidai wa daida ta karantawa da gyara abin da Shannon ya gabatar.

30–80
Kalmomin da ake nufi
5
Rabon gefe
0
Umarni mara kyau a T2I
1
Katin tabbatarwa da farko

01Me ya sa rubutu mai gudana ya fi alamomi: samfurin yana karatu kamar samfurin harshe

Tsofaffin samfuran diffusion da aka rubuta yawancin shawarwarin umarni domin su suna amfani da mai sanya rubutu wanda a aikace yake ɗaukar kansa kamar jakar tunani: nahawu bai kusan yin komai ba, kuma dabarar da ta yi nasara ita ce a tara sunaye masu alaƙa a ba su nauyi. Miyar alamomi ita ce siffar umarnin da ta dace da wannan mai sanyawa.

Samfurin Hoto na Shannon ba ya aiki haka. Yana karanta umarni kamar yadda samfurin harshe ke karanta rubutu, don haka tsarin nahawu yana ɗauke da ma'ana. "Tana riƙe laima a kafaɗarta" alaƙa ce tsakanin mutum, abu da wani ɓangaren jiki; "laima, kafaɗa" abubuwa biyu ne da suka faru suna nan. Kalmomin haɗi suna sanya abubuwa a wuri. Kalmomin aiki suna haifar da tsayuwa. Jimloli suna haɗa cikakken bayani da sunan da ya dace maimakon su bar shi ya shawagi a dukkan hoton.

Ga hoto iri ɗaya da aka nema ta hanyoyi biyu.

Salon alamomi — tsohuwar ɗabi'a
woman, 30s, tokyo, rain, night, crosswalk, neon signs, umbrella,
wet asphalt, reflections, masterpiece, best quality, 8k, ultra
detailed, (bokeh:1.3), cinematic lighting, photorealistic
Guntayen kalmomi 27 da aka raba da waƙafi. Babu kalmar aiki, babu alaƙa, kalmomi huɗu da suka mutu, tsarin nauyi guda ɗaya da ba a tantance ba.
Salon rubutu mai gudana — abin da ke aiki a nan
A woman in her early thirties waits at a rain-soaked Tokyo
crosswalk at night, holding a clear vinyl umbrella against her
shoulder and glancing down the street. Neon storefront signs
bleed red and cyan into the wet asphalt at her feet. Shot on
Kodak Portra 400, 85mm f/1.4, shallow depth of field, the
traffic signal behind her thrown out of focus.
Kalmomi 61. An sanya kowane abu dangane da wani, kuma an bayyana firam ɗin a matsayin shawarar kamara.

Ka duba abin da nau'in alamomi ba zai iya faɗi ba. Bai taɓa cewa an riƙe laimar ba, don haka ƙila ta ƙare tana kwance a ƙasa. Bai taɓa cewa neon yana haskakawa cikin kwalta ba, kawai cewa duka biyun suna nan, don haka ka samu alamomi da titi mai jiƙa ba tare da wata alaƙar haske tsakaninsu ba. Bai taɓa cewa ina matar take kallo ba, don haka tsayuwarta kwatsam ce. "Cinematic lighting" lakabin yanayi ne, ba tushen haske ba. Kuma kalmomi biyar na ƙarshe hayaniya ne, ana karanta su a matsayin kalmomin Turanci game da ɗaukar hoto maimakon riƙon sarrafawa.

Nau'in rubutu mai gudana ba ya tsawo don ya yi ado ba. Ya fi tsawo saboda ya amsa tambayoyin da nau'in alamomi ya bari a buɗe. Wannan shi ne dukkan dabarar.

02Tsawon umarnin hoto ya kamata ya kai nawa?

Ka nufi kalmomi 30 zuwa 80. Kada ka wuce 100 sosai. Wannan shi ne lever ɗin inganci mafi sauƙi da kake da shi, kuma bangarorin biyu na wannan faɗin suna da muhimmanci.

Ƙasa da kusan kalmomi 30 ba a bayyana tsari, haske da firam ba, don haka samfurin yana zaɓar duk abin da ya fi zama na yau da kullum a ƙididdiga. "Mace a shagon kofi" umarni ne da ya dace amma yana samar da hoto da za a manta, saboda ka miƙa kowace shawara mai ban sha'awa.

Sama da kusan kalmomi 100 wata gazawa daban tana bayyana: jimlolin baya suna gasa da na farko a kan yankuna iri ɗaya na hoton, kuma maimakon taruwar cikakken bayani sai ka samu narkewa. Bayanin haske na uku yana ɓata na farko, kuma batun da ka kafa a kalmomi takwas na farko yana samun ƙarancin kulawa da kowace jimla da ka ƙara. Dogayen umarni suna karantuwa kamar cikakku. Ba sa zanuwa a matsayin cikakku.

Ƙa'idar: ka rubuta umarnin da kake so, sannan ka rage shi zuwa jimlolin da za su canja hoton idan aka cire su.

03Ƙa'idar tsari: abu mafi muhimmanci da farko

Tunda samfurin yana karatu bi da bi, matsayi shi ne nauyi: abin da ya zo da farko yana kafa hoton, abin da ya zo ƙarshe yana daidaita shi. Ka sanya abubuwa cikin tsarin da samfurin yake buƙata:

  1. Batu — wa ko me wannan hoton yake nunawa, a takamaimai.
  2. Aiki ko tsayuwa — abin da suke yi, yadda aka riƙe su.
  3. Salo ko yanayi — hoton kamara, zanen mai, zanen fasaha, da matakin da yake ciki.
  4. Wuri da haske — inda wannan yake faruwa da wane haske ne yake faɗuwa a kansa.
  5. Fasaha — firam, ruwan tabarau, zurfin filin, takamaiman launuka.

Rauni da mutane suka fi yi wa kansu shi ne farawa da jimlar fasaha. Buɗewa da "Cinematic wide shot, anamorphic lens flare, teal and orange grade, of a man reading a newspaper" yana gaya wa samfurin cewa kamannin shi ne batun kuma mutumin ƙaramin bayani ne. Ka samu firam mai kyau, fanko, tare da ƙaramin siffa mara bayyana a ciki.

Fasaha da farko — batun ya iso a makare
Cinematic wide shot, anamorphic lens flare, teal and orange
color grade, shallow depth of field, golden hour, of a man
reading a newspaper on a park bench.
Batu da farko — abubuwa iri ɗaya, tsari daidai
A man in his sixties in a gray wool overcoat sits on a park
bench reading a folded broadsheet newspaper, one ankle crossed
over his knee. Late golden-hour sun rakes across him from the
left through bare plane trees. Photographed wide on a 35mm
anamorphic lens, shallow depth of field, warm highlights and
cool shadows.

Kowane abu daga nau'in farko ya tsira a na biyu; ba a yar da komai ba. Kawai yanzu samfurin ya san abin da hoton yake nunawa kafin a gaya masa yadda za a ɗauke shi — kuma wannan sake tsarin shi kaɗai yana motsa batun daga na kwatsam zuwa na tsakiya.

04Ka zama takamaimai game da batun

"Wani mutum" ba batu ba ne, wurin zama ne kawai — kuma samfurin yana cika wuraren zama da matsakaicin duk abin da ya gani, wanda shi ne ainihin inda kamanni mai sheƙi, mara shekaru, mara hali yake fitowa. Zama takamaimai yana ci kalmomi biyar kuma yana saya maka takamaiman mutum:

Mara bayyanaTakamaimai
wani mutumnamiji ɗan shekaru arba'in mai gemu mai launin gishiri-da-barkono da zurfin layukan murmushi
wata macemace ƴar ƙarshen shekaru ashirin mai gyaran gashi na undercut mai duhu da haƙorin gaba da ya karye
wani tsohon ginirumbun tubali mai hawa huɗu tare da alamar da aka fenta a kanta da tagogin sama da aka kafe
abincikwanon ramen tare da ƙwai mai laushi da aka raba biyu a sama, hayaƙi yana tashi

Ka lura da tsarin: shekaru ko zamani, siffa guda ta tsari, kuskure guda mai bambantawa. Abu na uku shi ne abin da ke sa sakamako ya ji kamar an ɗauke shi da kamara maimakon an ƙirƙira shi. Haƙori da ya karye, gefen tufafi da ya lalace, alamar da aka fenta a kanta — ƙananan rashin daidaito su ne alama mafi ƙarfi da ke hana kamannin filastik, kuma ba sa ci kusan komai a kasafin kalmominka.

05Hoto na gaske: ka rubuta buƙatar a matsayin hoton kamara

Hoto na gaske shi ne matakin da aka saba sai dai idan ka nemi fasaha, zane ko anime. Amma kalmar "photorealistic" ba ta yin aiki sosai. Abin da ke samar da hoton kamara shi ne bayyana aikin ɗaukar hoto: kamara ta gaske, ruwan tabarau na gaske, da nau'in fim ko tushen haske na gaske.

Dalilin a bayyane yake. Kowane hoto da aka rubuta masa "Kodak Portra 400" an yi shi ta wata takamaiman hanya mai takamaiman tsarin ƙwaya, sauƙin haske da nunin fata, kuma ambaton nau'in fim yana kiran duk hakan. "8k ultra detailed" ya taɓa bayyana ne kawai a shafukan sayar da hotuna, don haka yana kiran shafukan sayar da hotuna. Ƴan girke-girke masu aminci, kowanne yana samar da hoto daban na batu iri ɗaya:

Girke-girkeAbin da yake ba ka
Kodak Portra 400, 85mm f/1.4, soft window lightLaunin fata mai dumi mai gafara; rabuwa mai laushi; hoton mutum na mujalla
Ilford HP5 pushed to 1600, 35mm, available lightBaki da fari mai ƙwaya, mai bambanci, kuzarin takardar shaidar titi
Digital medium format, 80mm, single softbox at 45°Kamannin studio na kasuwanci mai tsafta, faɗuwar haske da aka sarrafa, matakin kayan sayarwa
135mm f/2, backlit at golden hour, atmospheric hazeBayan da aka matsa, hasken gefe, kamannin telephoto mai soyayya

Ka daidaita tushen bayani da girke-girken — siffar fata a matakin rami ba ta nufin komai a firam mai faɗi na 24mm inda fuskar tayi tsayin pixel tamanin — sannan ka ƙara shi. Cikakken bayani mai tushe shi ne shaidar siffa da ke nuna cewa ruwan tabarau ya ɗauki wannan da gaske. Ramukan fata da ake gani da ƙananan gashin fuska. Saƙar rigar lilin. Ƙurar da ke shawagi cikin katakon haske. Karcewa a takalmin fata. Gashin da ya tashi kan bango mai haske. Biyu ko uku sun isa, kuma su ne bambanci tsakanin zanen mutum da hoton kamara na mutum.

06Wurare masu shimfidu: gaba, tsakiya, baya

Idan wuri yana da zurfi, ka bayyana shimfidu uku daban kuma bisa tsari. Wannan ita ce dabara mafi tasiri ga hotunan da ya kamata su ji kamar wurare maimakon bangon baya, kuma kusan babu wanda yake yin ta. Ba tare da ita ba kana samun faɗi: duk abin da umarnin ya ambata yana sauka kusan a nesa ɗaya, yana gasa a kan tsakiyar firam ɗin. Bayyana shimfidu yana ba samfurin kasafin zurfi da zai kashe.

Shimfidu uku, an bayyana su a fili
In the foreground, a cracked concrete ledge with a half-empty
paper coffee cup and a scattering of cigarette ash, slightly out
of focus. In the midground, a bicycle courier in a yellow rain
jacket wheels her bike across a wet plaza, caught mid-stride. In
the background, a glass office tower dissolves into low fog.
Shot on a 50mm lens at f/2.8, flat overcast light.
Kalmomi 70. Abin gaba da ba a mai da hankali a kansa shi ne abin da ke sayar da zurfin — yana gaya wa samfurin inda kamarar take tsaye.

Gyare-gyare biyu. Ka sanya wani abu a gaba wanda da gangan ba a mai da hankali a kansa — hotunan gaske na wurare masu zurfi kusan koyaushe suna da ɗaya, kuma rashinsa alama ce da aka saba. Kuma ka faɗi wace shimfida batun yake ciki: idan mai kai kaya da keke shi ne manufar hoton yana cikin tsakiya, in ba haka ba hasumiya da kwafin kofi za su yi gasa da ita a firam.

07Ainihin launuka: lambobin hex, da aka haɗa da abubuwa

Ga aikin alama, ci gaba da kayan sawa, ko duk inda "ja" bai isa takamaimai ba, ka rubuta lambar hex kamar color #RRGGBB kuma ka haɗa ta da abu mai suna:

Daidai: hex da aka ɗaure da suna
...wearing a heavy wool peacoat, color #8B0000, over a cream
turtleneck, against a studio backdrop, color #1E3A5F.
Ba daidai ba: hex yana shawagi
...#8B0000, #1E3A5F, dark red and navy color palette, moody...
Babu abin da ke ɗaure waɗannan lambobin da komai, don haka suna aiki kamar shawarar launi mara bayyana a kan dukkan firam.

Ƙa'idodi uku masu amfani. Ka yi amfani da hex ga launuka biyu ko uku da suke da muhimmanci da gaske sannan ka yi amfani da kalmomi ga sauran — umarni mai lambobin hex shida yana kashe yawancin kasafinsa a kan ɗigon fenti. Ka ɗaure lambar da abu mai gefuna, kada ka ɗaure ta da "hasken" ko "yanayin". Kuma ka tuna cewa tana bayyana launin abu, ba ƙimar pixel na ƙarshe ba: rigar #8B0000 a hasken bayan lokacin zinare tana zanuwa da ɗumi fiye da samfurin launi, daidai kamar yadda za ta kasance a rayuwa.

08Zaɓar rabon gefe

Akwai rabo biyar kuma yawanci batun ne yake nuna wanda ya dace, amma buƙata bayyananna daga gare ka koyaushe ta fi ƙarfi — ka nemi bangon waya sai ka samu tall ba tare da la'akari da abin da abun cikin ke nunawa ba.

RaboYa fi dacewa daDalili
squareTsohon zaɓi. Abubuwa guda ɗaya a tsakiya, alamomi, tsari mai daidaito, saƙonnin sada zumuntaBabu babban gefe, don haka babu abin da firam ɗin ke yanke saboda karkatarsa
portraitMutane, cikakken jiki ko rabi; batutuwa masu tsawo kamar hasumiya, bishiyoyi, kwalabeYana ba wa mutum tsaye sarari daga kai zuwa ƙafa ba tare da rage shi ba
landscapeWurare, cikin gida, ƙungiyoyin mutane, hotunan kaya cikin mahalliSarari a kwance don wurin ya zama wuri da gaske
wideFiram na fim, hotunan fim, tutoci, hotunan kanun shafiZaɓin da ya fi faɗi; sarari fanko mai ƙarfi, jin anamorphic
tallFostoci, murfin littafi, bangon waya, tsarin sada zumunta na tsayeYa fi portrait tsawo; yana barin sarari da gangan don rubutu ko sama

Yin kuskure a nan ya fi lalata da yadda yake kama. Cikakken mutum tsaye a firam ɗin wide ba ya ba ka ƙaramin mutum a babban firam ba — yawanci yana ba ka wanda aka yanke, saboda samfurin yana cika faɗin sai kai ko ƙafa su fita daga firam. Ka daidaita firam da babban gefen batun sai yawancin ƙorafe-ƙorafen tsari su ɓace.

Halaye biyu da ya cancanci a tuna: gyara yana riƙe rabon gefe na hoton tushe ta atomatik, don haka ba za ka iya sake tsara firam ta gyara ba, yayin da haɗawa yana ba ka damar zaɓar sabon zane — sau da yawa hanya mafi tsafta ta canja siffar abin da ka riga ka mallaka.

09Munanan salo: abin da za ka daina yi

Kalmomin inganci: "masterpiece, best quality, 8k, ultra detailed"

Waɗannan riƙo ne da al'umma suka gano ga wani tsohon rabon horo takamaimai, inda wasu kalmomin bayani suka yi daidai da wasu tushen hotuna. Ba sa canjawa. A nan ana karanta su a matsayin kalmomin Turanci na yau da kullum, ba sa ba da gudunmawa, kuma suna ci wani ɓangare na kasafin kalmomin da aka riga aka gaya maka yana da ƙunci. Maimakon haka ka faɗi wane cikakken bayani kake so: ba "ultra detailed" ba amma "saƙar da ake gani a zaren rigar da ƙananan gemu a haɓa" — niyya iri ɗaya, an bayyana ta a matsayin abin da samfurin zai iya zanawa.

Tsarin nauyi: (word:1.3), [word], {{word}}

Ba a tantance su ba. Bakuna, ƙuloli, tsayayyun ɗigogi da lambobi suna sauka a matsayin rubutu na zahiri kuma, idan akwai wani abu, suna sa jimlar da ke kewaye ta yi wuyar karatu. Idan kana son ƙarin nauyi a kan wani abu, yi amfani da lever biyu da suke aiki: ka motsa shi gaba a cikin umarnin, kuma ka ba shi ƙarin kalmomi. "Mayafin siliki mai launin ja, gefunansa masu zare suna tashi cikin iska" ya fi "(red scarf:1.4)" nauyi ta kowace ma'auni da take da muhimmanci.

Hanyoyi masu saɓani

Haɗa umarni marasa jituwa ba ya haɗa su; yana samar da matsakaici mara ƙarfi wanda yawanci ba ya gamsar da ko ɗaya. Waɗanda suka fi yawa:

  • Hanya — "photorealistic" tare da "watercolor"; "oil painting" tare da "shot on 85mm". Ka zaɓi ɗaya ka tsaya a kansa.
  • Firam — "wide establishing shot" tare da "extreme close-up on her eyes"; "full body" tare da "detailed facial pores".
  • Haske — "harsh midday sun" tare da "soft golden hour glow"; "low key, deep shadows" tare da "bright and airy".
  • Zamani — "1970s film aesthetic" tare da "modern minimalist" tare da "cyberpunk".
  • Tara salo — masu fasaha huɗu ko nau'uka huɗu a lokaci ɗaya. Abubuwan dubawa biyu masu alaƙa na iya aiki; huɗu suna soke juna zuwa na gama-gari.
  • Yaɗuwa — "rustic, weathered, ancient, timeworn wooden door" tunani ɗaya ne da aka rubuta sau huɗu, kuma kowace maimaitawa tana narkar da sauran umarnin.

10Babu umarni mara kyau a rubutu-zuwa-hoto

Wannan shi ne munanan salon da ya fi kama ƙwararrun masu amfani, saboda a tsofaffin tsarin umarni mara kyau shi ne rabin fasahar.

AikiUmarni mara kyauAbin da za a yi
Rubutu-zuwa-hoto (ƙirƙira)Ba a goyon bayansaKa bayyana abin da kake so kawai. Kada ka taɓa rubuta "babu X".
Gyara hoto ɗayaAna goyon bayansaKa sanya abubuwan da ba a so a can, ka bar umarnin ya zama mai kyau.
Haɗa biyu ko fiyeAna goyon bayansaHaka nan — abubuwan da ba a so a mara kyau, rawa a mai kyau.

Ga ƙirƙira ƙa'idar cikakke ce kuma kaɗan ba a tsammani: rubuta "babu rubutu, babu alamar ruwa, babu ƙarin yatsu" a cikin umarnin yana sa waɗannan abubuwan su fi yiwuwa, ba ƙasa ba. Samfurin yana karanta "babu rubutu" a matsayin jimla mai ɗauke da tunanin rubutu; ƙi alama ce mai rauni kuma sunan mai ƙarfi ne. A zahiri ka nemi rubutu. Koyaushe ka faɗi nau'in mai kyau maimakon haka:

MaimakonKa rubuta
babu cunkoson bangobango mai launin toka mai tsafta marar ɗinki
ba mai ruɗi bamai kaifi sosai a idanu, f/8
babu wasu mutaneshi kaɗai a titi fanko da asuba
babu abubuwan zamanikicin na shekarun 1930 tare da kayan enamel da murhun ƙarfe
babu salon zane mai ban dariyashot on Kodak Portra 400, 85mm, natural skin texture

Ga gyara da haɗawa umarni mara kyau yana nan kuma ya kamata ka yi amfani da shi — amma ga abubuwan da ba a so, ba abun ciki ba. Kayan da suka dace: alamar ruwa, rubutu a saman hoto, gaɓoɓi da suka ninka, hannaye da suka karkace, da'irar kaifi fiye da kima, toshewar JPEG, layukan ɗinki a gefen haɗin. Kayan da ba su dace ba: duk abin da ya shafi abin da hoton yake nunawa, wanda yake cikin umarnin.

11Karanta umarnin da Shannon ya rubuta maka

Ga abin da ya sa wannan jagorar ta zama mai amfani nan take maimakon ta ka'ida. Ba dole ka rubuta umarnin ƙirƙira da kanka ba. Kana tambaya da harshe mai sauƙi, cikin kowane yare; Shannon yana gane cewa ana neman hoto, yana rubuta umarnin ƙirƙira na Turanci, sannan ya nuna maka shi a katin tabbatarwa kafin a zana komai. Ba a ƙirƙirar komai kuma ba a caje ka har sai ka tabbatar.

Don haka fasahar da ta fi biya ba ita ce rubuta umarni daga tushe ba — ita ce karanta wanda ke kan katin. Ka bi wannan jerin:

  1. Shin batun takamaimai ne? Idan katin ya ce "wani mutum" ko "wata mace", ka gano mafi girman ci gaban da ake da shi. Ka ba shi shekaru, girman jiki, wata siffa.
  2. Shin batun yana farko? Idan umarnin ya buɗe da salo ko kamara, ka nemi ya fara da batun maimakon haka.
  3. Shin yana cikin faɗin da ya dace? Ƙasa da kalmomi 30, ka nemi ƙarin takamaici kan haske da wuri. Sama da 100, ka nemi a rage shi.
  4. Shin akwai tushen haske? Ba kalmar yanayi ba — haske na gaske: taga, fitilar titi, walƙiyar kamara, sararin gizagizai, softbox.
  5. Akwai saɓani? Ka duba don hanyoyi biyu, firam biyu, ko yanayin haske biyu a umarni ɗaya.
  6. Akwai ƙi da aka ɓoye? Idan katin yana ɗauke da "ba tare da", "babu", ko "ka guji", wannan umarnin rubutu-zuwa-hoto ne da yake neman abin da ya ambata.
  7. Shin rabon ya dace da batun? Cikakken mutum a wide shi ne rashin daidaituwa da aka saba.

Gyare-gyare na hira ne: "ka sa ta ƴar farkon shekaru talatin mai tabo a gira ɗaya, ka ɗauke shi da Portra a 85mm, kuma ka yi amfani da portrait" umarni ne mai kyau sosai. Kana samun sabon kati sannan ka tabbatar da wancan maimakon. Babu hukunci kan maimaitawa, saboda ba a zana komai ba tukuna.

Kuma Shannon ba ya ƙirƙira ba tare da an nemi hakan ba. Gaisuwa, tambaya, ko hoton da ka haɗa don tattaunawa ba ya haifar da zanawa. Neman Shannon ya duba hoto — ya bayyana shi, ya karanta shi, ya yi nazarinsa — aikin gani ne na yau da kullum, ba ƙirƙira ba, kuma ba ya ci komai na ƙari.

12Yadda fasahar take sauyawa ga gyara da haɗawa

An rubuta ƙa'idodin da ke sama don ƙirƙira. Hanyoyi biyu makusanta suna canja siffar umarnin.

Gyara hoto ɗaya

Umarnin gyara umurni ne, ba bayani ba: ka faɗi me zai canja kuma me zai kasance. Bayyana sassan da ba za su canja ba shi ne mafi girman lever na inganci a samfurin — "Canja bayan zuwa titin birni cikin daren ruwan sama tare da hasken neon, ka bar matar kamar yadda take daidai" yana haɗuwa inda buƙata iri ɗaya ba tare da wannan jimlar ba tana sake zana fuskarta a shiru. Ka zama takamaimai game da angar: ka bar fuska, tsayuwa, gashi da tufafi ba tare da canji ba. Ana amfani da hoton tushe guda ɗaya kawai, kuma fitarwa tana riƙe rabon gefensa.

Haɗa hotuna biyu ko fiye

Duk lokacin da abubuwa suka fito daga hoto fiye da ɗaya — canja tufafi, abu cikin wuri, mutane biyu a firam ɗaya, kaya-a-kan-mutum, salo daga hoto ɗaya da batu daga wani — umarnin dole ya bayyana rawar kowane hoto, bisa tsari: "Matar daga hoto na farko sanye da rigar da aka saƙa mai ja daga hoto na biyu, tana tsaye a wannan wurin shakatawa." Hoto na farko shi ne babban batun ko wurin; hotuna na baya suna bayar da sassan da ake shigowa da su. Kada ka rage buƙatar hotuna da yawa zuwa tushe ɗaya tare da bayanin rubutu — haɗa hotunan na gaske shi ne abin da ke kiyaye ainihin kamannin waɗannan abubuwan.

13Umarni marar tantancewa: ka faɗi abin da kake nufi

Babu matatar abun ciki a hanyar ƙirƙira kuma babu shimfidar ƙi a kan samfurin da yake rubuta umarnin. Wannan shi ne dukkan manufar samfurin: janaretocin da aka tantance suna ƙin aiki na ƙwarai kullum — tsiraicin fasaha, zanen likita da na bincike na shari'a, hotunan binciken tsaro, tashin hankali cikin almara, izgilin siyasa, aikin alama da kama daga mai haƙƙin. Rabin fasahar umarni a dandalin da aka tantance injiniyan kalmomi masu laushi ne, kuma kalmomi masu laushi maƙiyin hoto mai daidaito ne.

Don haka akwai wani abu ɗaya da za a duba a katin: cewa yana bayyana ainihin batun da kalmomi bayyanannu maimakon kalmomin kauce wa batu da mutane suke koya daga kayan aikin da aka tantance. Zanen bincike na shari'a na rauni ya kamata ya faɗi menene raunin. Kalmomi masu kauce wa batu suna samar da hoto mafi muni a ko'ina, saboda samfurin yana zana abin da kalmomin suka faɗa.

Iyawa da hukunci abubuwa biyu ne daban. Shannon Lab LLC yana gudanar da wannan ƙarƙashin Manufar Amfani da Alhaki da aka buga: ƙirƙira marar matata kayan aiki ne na sana'a, kuma izini, haƙƙi da doka duk suna ci gaba da shafar abin da ka yi.

Jerin duba na shafi ɗaya

  • Rubutu mai gudana cikin cikakkun jimloli, ba alamomin da aka raba da waƙafi ba.
  • Kalmomi 30–80. Ka yanke duk abin da ba zai canja hoton ba.
  • Batu → aiki → salo → wuri da haske → fasaha.
  • Batu takamaimai: shekaru, siffa guda ta tsari, kuskure guda.
  • Hotunan kamara: ka ambaci kamara, ruwan tabarau, da nau'in fim ko tushen haske.
  • Wurare masu zurfi: gaba, tsakiya, baya, an bayyana su daban.
  • Ainihin launuka kamar color #RRGGBB da aka ɗaure da abu mai suna.
  • Ka zaɓi rabon daga babban gefen batun.
  • Siffofi biyu ko uku masu tushe, da suka dace da firam.
  • Babu kalmomin inganci, babu tsarin nauyi, babu hanyoyi masu saɓani.
  • Babu ƙi a rubutu-zuwa-hoto. Ƙi na gyara da haɗawa ne, kuma ga abubuwan da ba a so kawai.

14Inda API ya dace

Wani abu guda na daidaito ga masu karatu da suka zo daga bincike kan API: ƙirƙirar hoto, gyara da haɗawa suna cikin manhajar hira ta Shannon, ba a kan API na /v1 ba. API yana hidimar rubutu da gani — zai iya karantawa, bayyanawa da nazarin hotunan da ka aika masa a dukkan yaruka uku (/v1/chat/completions, /v1/messages, /v1/responses, duk masu streaming) — amma ba ya ƙirƙirar su. Idan kana gina tsarin da yake buƙatar fahimtar hotuna, takardun API su ne wurin farawa. Idan kana son yin hotuna, wannan manhajar hira ce, kuma duk abin da ke cikin wannan jagorar ya shafi can. Ana zana hotuna a gungun GPU namu.

15Tambayoyin da ake yawan yi

Tsawon umarnin hoto na AI ya kamata ya kai nawa?

Ka nufi kalmomi 30 zuwa 80 kuma kada ka wuce 100 sosai. Ƙasa da kalmomi 30 kana barin tsari, haske da firam ga kwatsam. Bayan kalmomi 100 jimlolin na baya suna fara gasa da na farko kuma cikakken bayani yana fara ɓacewa maimakon ya taru.

Ya kamata in rubuta alamomin kalmomi ko cikakkun jimloli?

Cikakkun jimloli. Samfurin Hoto na Shannon yana karanta umarni kamar yadda samfurin harshe ke karanta rubutu, don haka nahawu yana ɗauke da bayani. "Mace 'yar shekaru 30 a mashigin Tokyo mai ruwan sama da dare" tana samar da wuri mai haɗe; "mace, shekaru 30, Tokyo, ruwan sama, dare" tana samar da tarin siffofi marasa alaƙa da juna.

Zan iya amfani da umarni mara kyau don rubutu-zuwa-hoto?

A'a. Rubutu-zuwa-hoto a Shannon ba ya ɗaukar umarni mara kyau, don haka ka bayyana abin da kake so kawai kuma kada ka taɓa rubuta "babu X" a cikin umarnin — bayyana wani abu yakan kira shi. Gyaran hoto da haɗa hotuna suna goyon bayan umarni mara kyau, kuma can ne wurin da ya dace da abubuwan da ba a so kamar alamun ruwa, rubutu a saman hoto ko gaɓoɓi da suka ninka.

Shin kalmomi kamar "masterpiece, best quality, 8k" suna inganta hoton?

A'a. Waɗannan riƙo ne na tsohon zamanin samfuran kuma ba su ɗauke da wani bayani a nan — ana karanta su kamar kalmomi kawai, kuma suna ci wani ɓangare na kasafin kalmominka na 30–80 ba tare da faɗin komai ba. Haka nan ba a tantance tsarin nauyi kamar (word:1.3) ba; yana sauka kamar rubutu na zahiri. Maimakon haka ka bayyana ainihin cikakken bayanin da kake so.

Yaya zan samu ainihin launi?

Ka rubuta lambar hex kamar color #RRGGBB sannan ka haɗa ta da takamaiman abu mai suna, misali "a wool jacket, color #8B0000". Lambar hex da take shawagi ba tare da abu ba ba ta da abin da za ta ɗaure kanta da shi. Ka yi amfani da ita ga launuka biyu ko uku da suke da muhimmanci da gaske sannan ka bayyana sauran da kalmomi.

Zan iya ƙirƙirar hotuna ta hanyar API na Shannon?

A'a. Ƙirƙirar hoto, gyara da haɗawa suna faruwa a manhajar hira ta Shannon. API na /v1 yana hidimar rubutu da gani — zai iya karanta da nazarin hotunan da ka aika masa, amma ba ya ƙirƙirar su. Duk abin da ke cikin wannan jagorar ya shafi manhajar hira.

Gwada shi a kan umarni na gaske

Ka tambaya da kowane yare. Ka karanta katin. Ka daidaita shi. Sannan ka tabbatar.

Buɗe Hirar Shannon Ƙarin Bincike

Ba a zana komai, kuma ba a caje ka, har sai ka tabbatar da umarnin


Ƙarin karatu: Ƙirƙirar hoto ta AI marar tantancewa · Gyaran hoto na AI marar tantancewa · Manufar Amfani da Alhaki · Takardun API · Dukkan binciken Shannon. An ɗauko jagorar da ke nan daga gwajinmu na Samfurin Hoto na Shannon; shawarwarin da aka rubuta don tsofaffin tsarin diffusion galibi ba sa canjawa.

Dukkan hanyoyin bincike