Skip to content
DJEONAILWork with me
Semua tulisan
19 menit bacaJeon Nathaniel

Bedah Prompt Seedance 2.5: Cutscene ala GTA, 10 Shot dalam Satu Prompt

Satu prompt 10.697 karakter, 11 gambar referensi, 10 shot, dan 5 baris dialog dalam 30 detik. Ini bedah prompt Seedance 2.5 di balik video cutscene ala GTA: bagian yang dipatuhi model, yang meleset, dan cara mencobanya di Unzen.

seedancevideo aipromptunzen
Sampul: Bedah Prompt Seedance 2.5: Cutscene ala GTA, 10 Shot dalam Satu Prompt

Video 30 detik di bawah ini dikemas seperti cutscene game. Wira, tokoh utamanya, pamit dari tiga kawan bergaya GTA V, menangis di pelukan mereka, lalu berangkat dengan cabriolet merah bersama dua kawan bergaya GTA VI.

Gambarnya dibuat Seedance 2.5 dari satu prompt sepanjang 10.697 karakter dan 11 gambar referensi. Di artikel ini prompt itu dibedah per bagian, lengkap dengan apa yang dipatuhi model, apa yang meleset, dan cara mencobanya sendiri di Unzen.

Gambar dibuat Seedance 2.5 (30 detik, 1080p, 16:9). Subtitle ditambahkan saat editing.

Seedance 2.5 dalam satu menit

Seedance 2.5 adalah model video dari tim Seed ByteDance yang diumumkan 31 Juli 2026. Yang relevan untuk video ini:

  • Sampai 30 detik dalam sekali generate, naik dari 15 detik di Seedance 2.0, dan hasilnya bisa diperpanjang beberapa putaran.
  • Referensi jauh lebih banyak: sampai 30 gambar, 10 klip video, dan 10 klip audio dalam satu generate.
  • Gambar dan suara dibuat dalam satu proses, dan ByteDance menyebut keduanya tetap sinkron.
  • Beberapa shot dalam satu video. Model ini bisa menyusun beberapa shot yang saling terkait dalam 30 detik, dan prompt boleh memakai rentang waktu seperti 0-5 detik atau 6-10 detik.

ByteDance sendiri mengakui dua hal yang masih perlu diperbaiki: kewajaran fisik gerakan yang rumit, dan kestabilan adegan dengan banyak tokoh yang saling berinteraksi. Video ini berisi keduanya: pelukan, jabat tangan, dan enam tokoh.

Setelan yang dipakai, diambil dari catatan job render-nya: 30 detik, 1080p, 16:9, audio menyala, mode referensi dengan 11 gambar.

Anatomi prompt: 14 bagian

Prompt-nya ditulis dalam bahasa Inggris dan dibagi jadi 14 bagian berjudul huruf kapital. Hampir separuh isinya adalah TIMELINE.

Diagram panjang 14 bagian prompt: TIMELINE 4.633 karakter atau 43 persen, REFERENCE ROLES 1.433, AUDIO 780, bagian lain di bawah 500 karakter
Panjang tiap bagian prompt. TIMELINE sendiri 43 persen dari 10.697 karakter.

Pola besarnya: aturan yang berlaku untuk seluruh video ditulis sekali di bagian umum (FORMAT, CAMERA, LIGHTING, GRADE AND GRAIN, CONTINUITY, CONSTRAINTS), sedangkan yang berubah tiap detik ditulis di TIMELINE. Jadi tiap shot tidak perlu mengulang aturan cahaya atau grain.

FORMAT: semua angka di kalimat pertama

Kalimat pertamanya langsung memuat semua angka teknis:

30 seconds, 16:9 landscape composed for a central 4:3 safe area, 10 shots joined by hard cuts, natural real-time speed, no slow motion.

Durasi, rasio, area aman, jumlah shot, jenis transisi, dan kecepatan beres dalam satu baris. Kalimat berikutnya mengunci gaya: cutscene game generasi baru, grade golden hour peach, material fotoreal ala game engine. Lalu ada daftar larangan gaya: bukan kartun, bukan CGI plastik, bukan klip berita, bukan video vertikal.

REFERENCE ROLES: satu gambar, satu tugas

Setiap gambar referensi diberi satu tugas dengan pola yang sama, @ImageN controls only .... Lembar karakter hanya mengatur wajah, rambut, kulit, dan pakaian. Tiga kalimat pengaman diulang di bagian ini:

  • All four panels of @Image1 are one person, not four people. Lembar karakter berisi beberapa panel orang yang sama, dan kalimat ini mencegah model menganggapnya beberapa orang.
  • Do not copy pose, expression, studio backdrop, or lighting. Latar abu-abu studio di lembar karakter tidak boleh ikut masuk adegan.
  • Do not copy any background building from @Image9; place the car in @Image7/@Image8. Foto mobil hanya menyumbang mobilnya.

Di video, tidak ada latar studio abu-abu yang bocor, dan mobilnya muncul di jalan halaman depan, bukan di latar fotonya. Satu hal menarik: semua lembar karakter bergaya ilustrasi, tapi wajah di video keluar fotoreal, sesuai permintaan FORMAT soal wajah manusia hidup.

Papan 11 gambar referensi: lembar karakter Wira, Hassan, Raka, Banu, Lila, Seno, foto halaman depan, foto gerbang, foto cabriolet merah, dan dua sampul game parodi
Tugas tiap gambar referensi seperti tertulis di REFERENCE ROLES.

STARTING STATE: isi frame pertama

Bagian ini menggambarkan detik nol: Hassan menyodorkan sampul pertama, bahu Wira terlihat di tepi kiri, Raka dan Banu ada di dekat situ tapi belum masuk frame. Kalimat yang paling berguna justru soal benda yang belum terlihat:

The 2nd disk case already lies on the car bed, unseen until Shot 8.

Prop untuk shot 8 sudah diletakkan sejak frame pertama, bukan muncul tiba-tiba di tengah video.

TIMELINE: 10 shot dengan stopwatch

Setiap shot ditulis dengan urutan yang sama: rentang detik, ukuran shot, tinggi kamera, fokus, gerak kamera, aksi, dialog lengkap dengan detik dan nada bicaranya, lalu titik potong. Contohnya shot 1:

Shot 1, 0.00-2.54s. Close-up product insert matching @Image10. Eye level, slightly high to the case, very shallow DOF, circular courtyard bokeh. Soft handheld or very slow push. Hassan's right linen-sleeved hand offers the first disk @image10 case, title sharp at center. [...] 0.00-0.80s: Hassan says in calm solemn Indonesian: {dua tahun...} Pause. 1.32-2.40s: {kamu pantas dapat ini.} Soft musical sting near 0.20s. Cut while the case is still offered.

Tiga pola yang dipakai di seluruh timeline:

  • Titik potong ditulis sebagai momen, bukan cuma angka. Misalnya "Cut while the case is still offered", "Cut on that stillness", dan "Cut at the peak of the emotion".
  • Dialog dikunci ke pemiliknya. Shot berdialog ditutup kalimat seperti Only Raka's mouth moves. atau Wira's mouth stays closed.
  • Dialog ditulis di dalam kurung kurawal, misalnya {kamu pantas dapat ini.}, sehingga terpisah jelas dari deskripsi adegan.

CAMERA, LIGHTING, GRADE AND GRAIN: aturan untuk semua shot

  • CAMERA berisi daftar gerakan yang boleh, per shot. Shot lain dikunci (Every other shot is locked). Lensa diatur per ukuran shot: sekitar 85mm untuk close-up, 35-50mm untuk medium, 24-35mm untuk wide. Zoom, whip pan, crane, dan drone dilarang.
  • LIGHTING hanya punya satu sumber cahaya: matahari sore yang rendah. Flare hanya boleh muncul di shot 4 dari kanan, dan di video flare itu memang muncul di shot 4.
  • GRADE AND GRAIN meminta warna dan grain yang identik di setiap cut. Grade peach-nya tidak berubah dari shot pertama sampai terakhir.

SKIN AND TEXTURE dan PHYSICS: melawan wajah plastik

Dua bagian pendek ini meminta detail yang sering hilang di video AI: pori, bulu halus yang tertangkap cahaya, jejak air mata sungguhan di shot 4, 6, dan 9, serta tekstur kain dan serat kayu. PHYSICS meminta air mata turun mengikuti gravitasi, tangan berjari lima, dan sampul yang terasa punya berat. Di video, air mata terlihat jelas di shot 4 dan 6, dan di shot 9 Wira mengusap matanya dengan lengan jas.

PERFORMANCE dan AUDIO: siapa bicara, kapan

PERFORMANCE memberi tiap tokoh satu arah emosi. Wira tenang, lalu sedih, lalu mantap. Hassan khidmat dan bangga, Banu kasar tapi setia. Aturan terpentingnya: mulut hanya bergerak saat tokoh itu bicara, dan setelah detik 17,08 tidak ada lagi yang bicara.

AUDIO mengunci lima baris dialog berbahasa Indonesia, warna suara tiap tokoh, tiga efek suara (gesekan kain saat jabat tangan, langkah kaki ke bumper, ketukan sampul), dan musik instrumental sekitar 120 BPM yang mengecil saat ada dialog. Kalimat terakhirnya penting:

This generated audio is a timing scratch only; finished voices and music will be replaced in post.

Artinya audio dari model diperlakukan sebagai panduan timing, dan suara final memang direncanakan diganti saat editing.

CONTINUITY, ENDING STATE, CONSTRAINTS: pagar terakhir

CONTINUITY melacak prop. Sampul pertama harus tetap di tangan Wira dari akhir shot 1 sampai shot 8, lalu diletakkan di samping sampul kedua. Di video, alurnya utuh: sampul itu terlihat di tangannya saat berpelukan di shot 4, disodorkan ke kru baru di shot 7, dan diletakkan di shot 8. ENDING STATE menggambarkan frame terakhir, dan CONSTRAINTS menutup dengan area aman 4:3, sampul yang harus persis seperti referensi, serta larangan caption dan watermark.

Rencana vs hasil: urutannya tepat, detiknya tidak

Kesepuluh shot keluar dengan urutan yang sama persis dengan TIMELINE. Yang bergeser adalah durasinya.

Perbandingan timeline: 10 shot rencana dan 10 shot video final berurutan sama, shot 1 dan 2 lebih panjang, shot 7 sampai 9 lebih pendek
Garis kuning menghubungkan awal shot yang sama di rencana dan di video final.

Shot 1 baru berakhir di detik 3,6, bukan 2,54. Shot 2 juga lebih panjang, sehingga awal shot 3 terlambat 2,3 detik dari rencana. Di bagian akhir, shot 7 sampai 9 dipadatkan dari 8,5 jadi 5,5 detik, dan shot 10 malah mulai lebih awal. Catatan: video final ini sudah melewati editing dan panjangnya 29,4 detik, jadi selisih kecil bisa juga berasal dari editing.

Yang meleset, dan kenapa itu berguna

Empat temuan: plat nomor ikut gambar bukan teks, pakaian Hassan ikut teks bukan lembar karakter, papan kayu muncul dari sisa kata pickup, area aman 4:3 dilanggar di shot 2
Teks prompt, gambar referensi, dan video final dibandingkan berdampingan.
  1. Plat nomor ikut gambar. Prompt minta "GG Osl0" dengan angka nol. Foto mobil di @Image9 berplat GG OSLO, dan plat itulah yang keluar di video.
  2. Pakaian ikut teks. Lembar karakter Hassan memakai jas abu-abu, sedangkan teks prompt menyebut jas linen krem. Di video, wajahnya ikut lembar karakter dan jasnya ikut teks. Raka juga begitu: lembarnya kemeja biru, di video overshirt olive sesuai teks. Jadi model tidak selalu memihak sumber yang sama. Cara paling aman adalah menyamakan isi teks dan gambar.
  3. Sisa kata dari versi lama. Prompt masih menyebut "red pickup", "wood-slat bed", "cab", dan "roll bar", padahal @Image9 sebuah cabriolet. Hasilnya, di shot 8 model membuatkan papan kayu tempat sampul diletakkan, sesuatu yang tidak dimiliki cabriolet. Setiap kali referensi diganti, cari kata-kata versi lama di seluruh prompt.
  4. Area aman 4:3 dilanggar di shot 2. CONSTRAINTS minta 12,5 persen kiri dan kanan frame bebas dari info penting. Di shot 2, kepala Wira dan wajah Raka masuk ke area itu, jadi versi 4:3 akan memotong mereka.
  5. Detail kecil sampul berubah. Sampul @Image10 saat diletakkan di shot 8 nyaris persis, tapi label rating "RP" di referensi berubah jadi "18", dan muncul teks "S15" yang tidak ada di referensi. Judul besar aman, teks kecil tetap perlu dicek.

Ceklis sebelum generate prompt panjang

Prompt lengkap

Ini prompt yang dipakai untuk render final, tanpa diubah, termasuk typo dan sisa kata pickup-nya. Tag @Image1 sampai @Image11 merujuk gambar referensi sesuai urutannya.

FORMAT
30 seconds, 16:9 landscape composed for a central 4:3 safe area, 10 shots joined by hard cuts, natural real-time speed, no slow motion.
An original cinematic farewell cutscene in a next-generation story-game visual language: warm neon-peach golden-hour grade, game-engine photoreal materials, cutscene camera language, spherical bokeh, filmic highlight roll-off. Living human faces must match the photoreal references. Not a cartoon, not plastic CGI, not a news clip, not a vertical social video.

REFERENCE ROLES
@Image1 controls only Wira, a young man saying goodbye to close friends: face, hair, glasses, skin, wardrobe as shown. All four panels of @Image1 are one person, not four people. Do not copy pose, expression, studio backdrop, or lighting.
@Image2 controls only Hassan, the older mentor in the sand-beige linen suit, same rule: face, hair, skin, wardrobe; one person across all panels; no pose/backdrop copy.
@Image3 controls only Raka, the friend in the olive overshirt, same rule.
@Image4 controls only Banu, the burly loyalist in the olive-drab tank, including his unlettered tattoos, same rule.
@Image5 controls only Lila, a friend from Wira's next chapter, with the short dark-brown bob and unlettered hibiscus tattoo, same rule.
@Image6 controls only Seno, a friend from Wira's next chapter, in the pale mint camp shirt with unlettered arm tattoos, same rule.
@Image7 controls only the courtyard architecture of the Palemputih grounds, a fictional private estate: white classical portico, clipped rounded trees, circular lawn, flagpole, curved asphalt drive, peach sky. Do not add or remove buildings. Invented place; not a real landmark or public office.
@Image8 controls only the reverse view toward the white-and-iron front gate, masonry posts, palisade and driveway, used in the ending wide.
@Image9 controls only the red classic Cabriolet. Do not copy any background building from @Image9; place the car in @Image7/@Image8.


STARTING STATE
Golden hour, courtyard from @Image7. Hassan offers the first disk @image10 case across a pale limestone ledge. Wira's white-shirt left shoulder is at the left edge, face not yet visible. Raka and Banu are nearby, not yet in frame. The red pickup is out of frame. The 2nd disk case already lies on the car bed, unseen until Shot 8. Camera is close on the case, slightly high, 85mm-look, very shallow depth of field.

TIMELINE
Shot 1, 0.00-2.54s. Close-up product insert matching @Image10. Eye level, slightly high to the case, very shallow DOF, circular courtyard bokeh. Soft handheld or very slow push. Hassan's right linen-sleeved hand offers the first disk @image10 case, title sharp at center. Wira's left shoulder stays an out-of-focus silhouette;. 0.00-0.80s: Hassan says in calm solemn Indonesian: {dua tahun...} Pause. 1.32-2.40s: {kamu pantas dapat ini.} Soft musical sting near 0.20s. Cut while the case is still offered.

Shot 2, 2.54-6.58s. Starts as a medium 3-shot over Wira's shoulder: Wira now holds the case, Hassan center, Raka right, Palemputih courtyard behind. One continuous take: gentle pan/push into a medium-close hug, then pull-back to the medium 3-shot as they release. Eye level. At 3.86s Raka says in warm casual Indonesian: {Denger, bro, jangan lupa sembilan belas juta satu lapangan pekerjaan.} They shake hands, hug firmly, release and shake again. Hassan smiles; he may leave the tight frame during the hug and return on the pull-back. Only Raka's mouth moves. Cut after the second handshake settles.

Shot 3, 6.58-9.33s. Medium-close over-the-shoulder. Banu is sharp; Wira is a soft foreground silhouette at left; sunset estate and trees behind. Almost locked, tiny drift, eye level. Banu brings his hands together before his mouth in a quiet, respectful gesture of thanks. Eyes open at first, then close, expression more pained, a slight bow. He does not speak. Quiet courtyard air and a very low underscore only. Cut on that stillness.

Shot 4, 9.33-12.96s. Close-up hug. Wira on the left, eyes closed, crying; Banu's back and head on the right. Tight, shallow, warm sun flare from the right. Mid-shot the frame tightens and the forst disk case appears in Wira's hand against Banu's back. At 10.40s Banu says in mid, intense, caring Indonesian: {Kalau mereka ga enak sama kamu, telepon aku.} Wira's mouth stays closed. Cut at the peak of the emotion.

Shot 5, 12.96-14.79s. Wide from behind the red car cabriolet on the courtyard drive, grounds and distant gate ahead. Lower-to-medium height, almost locked, deeper focus. Lila and Seno in the cab turn back, smiling; Seno raises a full wave. Contrast cut from tears to joy. From the first frames Lila calls in excited, almost-shouted Indonesian: {Ayo, kita udah nunggu lama banget!} Only Lila's mouth moves. No dialogue after 14.56s.

Shot 6, 14.79-18.75s. Close-up portrait of Wira, head and shoulders, courtyard bokeh. Eye level, shallow, warm key from the right, practically locked. 14.79-15.50s: bittersweet watery-eyed smile to lens. 15.50-17.08s: Wira says in soft grateful Indonesian: {Makasih, kawan, untuk semuanya.} 17.08-18.00s: calmer smile, no speech. 18.30-18.75s: he dips his head, turns right, starts walking, almost leaving frame. No more dialogue after 17.08s in the film. Cut as he turns away.

Shot 7, 18.75-21.17s. Wide matching Shot 5, same axis and height, locked. Wira enters from frame left, walks to the red car cabriolet, climbs into the wood-slat bed by the bumper, and holds out the first disk @image10 case to Lila and Seno, who turn and smile. No dialogue. Underscore plus soft footsteps and a shoe-on-metal bumper step. Cut as he is on the bed with the case in hand.

Shot 8, 21.17-23.67s. Overhead insert, almost locked, looking down at the worn wood car bed. The second disk @image11 case already lies on the left, title readable. Wira's hand enters from the right, sets the first disk @image10 case neatly beside it, title readable, aligns it, then lifts away. No dialogue. Underscore plus a soft case-shell tap. Cut on the two cases side by side.

Shot 9, 23.67-27.25s. Medium / medium-close. Wira sits on the car bed, knees up, facing back toward the three mentors (toward camera). Behind him: Seno and Lila in the cab, the roll bar, estate grounds. Eye level slightly low, relatively locked. 23.67-24.90s: he wipes his eyes or face with a sleeve. A small open-mouth breath around 25.4s that is not speech, then a gentle smile until the cut. No dialogue. Underscore only. Cut on that look-back smile.

Shot 10, 27.25-30.00s. Wide from behind. Banu left, Hassan center (left hand on Banu's shoulder, right hand waving), Raka right waving, all in the foreground. The red car cabriolet with visible License plate "GG Osl0" recedes down the drive toward the white-and-iron gate from @Image8 under a peach sky. Locked or with an almost imperceptible slow push that does not enlarge the three men. They stand and wave. Banu shifts slightly. No dialogue. Underscore softens. End on the car moving toward the gate, not on a return to the cases.

CAMERA
Only these moves: Shot 1 soft handheld or slow push to the case; Shot 2 pan/push into the hug and pull-back; Shot 3 tiny drift; Shot 4 slight mid-shot tightening onto the case; Shot 8 motion led by the placing hand; Shot 10 optional almost imperceptible push. Every other shot is locked. About 85mm on close-ups, 35-50mm on mediums, 24-35mm on wides, spherical bokeh, shallow DOF on close-ups, deeper on wides. No zooms, whip pans, crane, or drone.

LIGHTING
One motivated source: late golden-hour sun, peach-amber, long soft shadows, gentle bloom, warmer rim on linen and hair, slightly cooler shade. No ring light, no beauty fill. A warm sun flare is allowed only in Shot 4, from the right. Lighting ratio stays consistent across every cut.

SKIN AND TEXTURE
Living skin: pores, vellus hair in the rim light, uneven tone. In Shots 4, 6 and 9, Wira has wet lash lines and real tear tracks, not beauty-filter crying. Real fabric weave, real case-shell wear, real wood grain on the bed. No plastic skin, no over-white teeth, no glowing eyes.

PHYSICS
Tears run with gravity. Hair moves with breath. Hands have five correct fingers and a real grip. The case has weight. The car recedes with real perspective. Nothing floats or morphs.

GRADE AND GRAIN
Warm neon-peach golden hour with next-gen cutscene polish: lifted but not crushed blacks, soft highlight roll-off, gentle bloom, fine 35mm-like grain identical on every shot. Grade does not change across cuts.

PERFORMANCE
Wira leaves with dignity and gratitude: still, then openly sad, then quietly resolved. Hassan is solemn and proud. Raka is warm and casual. Banu is rough and deeply loyal, almost prayerful, then protective in the hug. Lila is joyful and kindly impatient. Seno is easy, waving from the cab. Mouths move only during that character's own lines. After 17.08s nobody speaks.

CONTINUITY
Same six faces, hair, wardrobe, props, estate, car and light in every shot. The first disk @image10 case stays with Wira from the end of Shot 1 until Shot 8, when it is placed beside the second disk @image11 case already on the bed; both stay there through Shots 9-10. No lettering on tattoos,. Architecture always matches @Image7 and @Image8, never the extra building in @Image9.

AUDIO
Clean lip-synced casual Indonesian, exactly the five written lines, no added or repeated lines. Hassan: low, calm, solemn. Raka: low, warm, casual. Banu: mid, intense, caring. Lila: high, excited, almost called-out. Wira: high-mid male, soft, grateful. Room tone: quiet estate courtyard at dusk. Practicals only: handshake cloth in Shot 2, footsteps and bumper step in Shot 7, case tap in Shot 8. Score: original instrumental cinematic farewell, no lyrics, about 120 BPM with a half-time feel in the first 8 seconds, a soft sting near 0.20s, ducked under dialogue until 17.08s, then carrying the last 13 seconds and softening at the end. No narrator, no English dialogue, no crowd. This generated audio is a timing scratch only; finished voices and music will be replaced in post.

ENDING STATE
Banu, Hassan and Raka stand in the foreground, still waving. The red car cabriolet is farther down the drive, heading toward the white-and-iron gate under a peach sky. The two cases are on the bed, no longer readable at this distance. Last sound: softening underscore, no dialogue.

CONSTRAINTS
Compose every shot so all faces, hands and props sit inside the central 4:3 safe area; keep the far left and far right 12.5 percent of the 16:9 frame free of critical information for a later 4:3 crop. Disk case cover @image10 and @image11 must exactly the same as reference
No captions, subtitles, watermarks or lower-thirds.

Mencoba Seedance 2.5 di Unzen

Seedance 2.5 bisa dicoba di Unzen. Langkahnya per 17 November 2026:

  1. Masuk ke Unzen, buka Studio, lalu pilih Video. Mode Video butuh paket berbayar.
  2. Pilih model Seedance 2.5 R2V, yaitu mode referensi.
  3. Isi Start image (wajib) dan Reference images (wajib, maksimal 9). Audio referensi boleh ditambahkan.
  4. Tempel prompt. Batasnya 15.000 karakter, jadi prompt 10.697 karakter di atas muat.
  5. Atur durasi (4 sampai 30 detik), resolusi (480p, 720p, atau 1080p), rasio (misalnya 16:9), dan audio.
  6. Kirim, lalu tunggu hasilnya muncul di Studio.

Coba Seedance 2.5 di Unzen

Sumber

Fakta di artikel ini dicek pada 17 November 2026.

  • ByteDance Seed: One-take Creation, Flexible Referencing: Introducing Seedance 2.5 (31 Juli 2026), sumber durasi, batas referensi, kemampuan multi-shot, dan kelemahan yang diakui.
  • Setelan render (durasi, resolusi, rasio, jumlah referensi) dari catatan job render video ini. Waktu potong dari deteksi pergantian adegan ffmpeg pada video final.
  • Batas Seedance 2.5 R2V di Unzen dari katalog model dan tampilan Studio Unzen.
Newsletter

Mau dapat kabar tiap ada tulisan baru?

Subscribe newsletter gratis dengan nama, email, dan nomor WhatsApp. Tulisan dan panduan baru akan dikabarkan ke kamu. Tanpa daftar akun.

Sesekali saja, dan bisa berhenti kapan pun

Ada yang mau ditanyakan soal tulisan ini? Kirim email.