NotePublik Pro

{
    "env": {
        "ANTHROPIC_BASE_URL": "http://localhost:20128/v1",
        "ANTHROPIC_AUTH_TOKEN": "sk-1b84bd13d72c0be8-j9y0vs-0f0d00d9",
        "ANTHROPIC_MODEL": "sonnet[1m]",
        "ANTHROPIC_DEFAULT_HAIKU_MODEL": "gravity[1m]",
        "ANTHROPIC_DEFAULT_SONNET_MODEL": "gravity[1m]",
        "ANTHROPIC_DEFAULT_OPUS_MODEL": "gravity[1m]",
        "CLAUDE_CODE_SUBAGENT_MODEL": "sonnet[1m]",
        "MAX_THINKING_TOKENS": "1000000",
        "CLAUDE_CODE_EFFORT_LEVEL": "xhigh"
    },
    "model": "sonnet",
    "effortLevel": "xhigh",
    "ultracode": true,
    "autoUpdatesChannel": "latest",
    "hasCompletedOnboarding": true
}

# RunningHub OpenAPI allows you to call AI applications and workflows via HTTP requests. ## Start Using ### Register User After registering a RunningHub account and topping up your wallet, you can start using the AI Application API and ComfyUI Workflow API. Note: If you're using a Consumer-Membership API Key, you must have a Personal membership or higher to access these APIs. Users with Enterprise Shared or Enterprise Dedicated API Keys are not subject to this restriction. ### Get Your API Key RunningHub automatically generates a unique 32-bit API KEY for each user Please keep your API KEY safe and do not disclose it. Subsequent steps will depend on this key for operations ### Submit a request Submit an API request. RunningHub API has handled the API Key for you, you just need to submit the request ```curl curl --location --request POST 'https://www.runninghub.ai/openapi/v2/run/workflow/2085385545848819714&#039; \ --header "Content-Type: application/json" \ --header "Authorization: Bearer ${RUNNINGHUB_API_KEY}" \ --data-raw '{ "addMetadata": true, "nodeInfoList": [], "instanceType": "default", "usePersonalQueue": "false" }' ``` #### Request Parameters | Schema | Type | Required/Optional | The generated result from the AI application. | | --- | --- | --- | --- | | `addMetadata` | Boolean | Optional | Whether to add workflow metadata to output images | | `nodeInfoList` | List | Required | List of node parameter mappings used to dynamically modify workflow parameters | | `instanceType` | String | Optional | Specify the running instance type<br>default (24G VRAM), plus (48G VRAM) | | `usePersonalQueue` | Boolean | Optional | Whether to use a personal exclusive queue | | `retainSeconds` | Integer | Optional | Instance retention duration in seconds. Applies only to Enterprise Shared API Keys. After a successful task, the same user's same workflow can preferentially reuse the retained instance within this window to reduce cold starts and queueing. Additional charges apply for the retention period and are billed by actual retained duration. Valid range: 10-180 seconds. | | `webhookUrl` | String | Optional | Webhook callback URL. RunningHub will send a POST request when the task completes | #### Response Example ```json { "taskId": "2013508786110730241", "status": "RUNNING", "errorCode": "", "errorMessage": "", "results": null, "clientId": "f828b9af25161bc066ef152db7b29ccc", "promptTips": "{\"result\": true, \"error\": null, \"outputs_to_execute\": [\"4\"], \"node_errors\": {}}" } ``` #### Response Fields | Schema | Type | The generated result from the AI application. | | --- | --- | --- | | `taskId` | String | Task ID, used to query task status later | | `status` | String | Current task status. Common states: QUEUED, RUNNING, SUCCESS, FAILED | | `errorCode` | String | Error code, returned only upon failure | | `errorMessage` | String | Detailed error message | | `results` | List | Generation results (null when submitting) | | ├ `url` | String | Important: This link is valid for only 24 hours. After task generation is complete, be sure to download the video file or transfer it to your server within this time window. Once expired, the link will be permanently invalid and cannot be restored. | | ├ `nodeId` | String | Workflow node ID that produced this result | | ├ `outputType` | String | File extension (e.g., png, mp4, txt) | | └ `text` | String | Content is displayed here if the output is plain text | | `clientId` | String | Client session ID, used to identify the current connection | | `promptTips` | String (JSON) | Validation info from the ComfyUI backend, including node IDs to execute or debugging info | ### Query Result Query the execution progress and final results by Task ID #### Request Example ```curl curl --location --request POST 'https://www.runninghub.ai/openapi/v2/query&#039; \ --header "Content-Type: application/json" \ --header "Authorization: Bearer ${RUNNINGHUB_API_KEY}" \ --data-raw '{ "taskId": "${RUNNINGHUB_TASKID}" }' ``` #### Response Example ```json { "taskId": "2013508786110730241", "status": "SUCCESS", "errorCode": "", "errorMessage": "", "failedReason": {}, "usage": { "consumeMoney": null, "consumeCoins": null, "taskCostTime": "0", "thirdPartyConsumeMoney": null }, "results": [ { "url": "https://rh-images-1252422369.cos.ap-beijing.myqcloud.com/b04e28cad0ee39193921a30a2eb4dc00/output/ComfyUI_00001_plhjr_1768892915.png&quot;, "nodeId": "2", "outputType": "png", "text": null } ], "clientId": "", "promptTips": "" } ``` #### Response Fields | Schema | Type | The generated result from the AI application. | | --- | --- | --- | | `taskId` | String | Task ID | | `status` | String | Final task status. SUCCESS indicates successful generation | | `results` | List | List of generation results, including images, videos, or text outputs | | ├ `url` | String | Important: This link is valid for only 24 hours. After task generation is complete, be sure to download the video file or transfer it to your server within this time window. Once expired, the link will be permanently invalid and cannot be restored. | | ├ `nodeId` | String | Workflow node ID that produced this result | | ├ `outputType` | String | File extension (e.g., png, mp4, txt) | | └ `text` | String | Content is displayed here if the output is plain text | | `errorCode` | String | Error code (if any) | | `errorMessage` | String | Error message (if any) | | `failedReason` | Object | ComfyUI-related failure reason | | `usage` | Object | Task usage information | | ├ `thirdPartyConsumeMoney` | String | Third-party API consumed amount | | ├ `consumeMoney` | String | Runtime duration cost | | ├ `consumeCoins` | String | RH coins consumed | | └ `taskCostTime` | String | Execution time (ComfyUI workflow runtime) | ### File Upload Resource file (eg: `imageUrls`) parameters accept either public file URLs or Base64 data URIs. #### Public URL Pass a publicly accessible URL directly: ```json { "imageUrls": [ "https://example.com/image.png&quot; ] } ``` #### Base64 data URI Embed images in Base64 format: ```json { "images": [ "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA..." ] } ``` #### RH Upload API Upload a local file to obtain a URL. **Endpoint:** `https://www.runninghub.ai/openapi/v2/media/upload/binary` **Request** ```curl curl --location --request POST 'https://www.runninghub.ai/openapi/v2/media/upload/binary&#039; \ --header 'Authorization: Bearer [Your API KEY]' \ --form 'file=@/path/to/image.png' ``` **Response** ```json { "code": 0, "message": "success", "data": { "type": "image", "download_url": "xxxx.png", "fileName": "openapi/xxxx.png", "size": "3490" } } ``` **Remark:** The uploaded URL is valid for one day. After expiration, it cannot be accessed directly by URL.

# RunningHub OpenAPI allows you to call AI applications and workflows via HTTP requests. ## Start Using ### Register User After registering a RunningHub account and topping up your wallet, you can start using the AI Application API and ComfyUI Workflow API. Note: If you're using a Consumer-Membership API Key, you must have a Personal membership or higher to access these APIs. Users with Enterprise Shared or Enterprise Dedicated API Keys are not subject to this restriction. ### Get Your API Key RunningHub automatically generates a unique 32-bit API KEY for each user Please keep your API KEY safe and do not disclose it. Subsequent steps will depend on this key for operations ### Submit a request Submit an API request. RunningHub API has handled the API Key for you, you just need to submit the request ```curl curl --location --request POST 'https://www.runninghub.ai/openapi/v2/run/workflow/2085385545848819714&#039; \ --header "Content-Type: application/json" \ --header "Authorization: Bearer ${RUNNINGHUB_API_KEY}" \ --data-raw '{ "addMetadata": true, "nodeInfoList": [], "instanceType": "default", "usePersonalQueue": "false" }' ``` #### Request Parameters | Schema | Type | Required/Optional | The generated result from the AI application. | | --- | --- | --- | --- | | `addMetadata` | Boolean | Optional | Whether to add workflow metadata to output images | | `nodeInfoList` | List | Required | List of node parameter mappings used to dynamically modify workflow parameters | | `instanceType` | String | Optional | Specify the running instance type<br>default (24G VRAM), plus (48G VRAM) | | `usePersonalQueue` | Boolean | Optional | Whether to use a personal exclusive queue | | `retainSeconds` | Integer | Optional | Instance retention duration in seconds. Applies only to Enterprise Shared API Keys. After a successful task, the same user's same workflow can preferentially reuse the retained instance within this window to reduce cold starts and queueing. Additional charges apply for the retention period and are billed by actual retained duration. Valid range: 10-180 seconds. | | `webhookUrl` | String | Optional | Webhook callback URL. RunningHub will send a POST request when the task completes | #### Response Example ```json { "taskId": "2013508786110730241", "status": "RUNNING", "errorCode": "", "errorMessage": "", "results": null, "clientId": "f828b9af25161bc066ef152db7b29ccc", "promptTips": "{\"result\": true, \"error\": null, \"outputs_to_execute\": [\"4\"], \"node_errors\": {}}" } ``` #### Response Fields | Schema | Type | The generated result from the AI application. | | --- | --- | --- | | `taskId` | String | Task ID, used to query task status later | | `status` | String | Current task status. Common states: QUEUED, RUNNING, SUCCESS, FAILED | | `errorCode` | String | Error code, returned only upon failure | | `errorMessage` | String | Detailed error message | | `results` | List | Generation results (null when submitting) | | ├ `url` | String | Important: This link is valid for only 24 hours. After task generation is complete, be sure to download the video file or transfer it to your server within this time window. Once expired, the link will be permanently invalid and cannot be restored. | | ├ `nodeId` | String | Workflow node ID that produced this result | | ├ `outputType` | String | File extension (e.g., png, mp4, txt) | | └ `text` | String | Content is displayed here if the output is plain text | | `clientId` | String | Client session ID, used to identify the current connection | | `promptTips` | String (JSON) | Validation info from the ComfyUI backend, including node IDs to execute or debugging info | ### Query Result Query the execution progress and final results by Task ID #### Request Example ```curl curl --location --request POST 'https://www.runninghub.ai/openapi/v2/query&#039; \ --header "Content-Type: application/json" \ --header "Authorization: Bearer ${RUNNINGHUB_API_KEY}" \ --data-raw '{ "taskId": "${RUNNINGHUB_TASKID}" }' ``` #### Response Example ```json { "taskId": "2013508786110730241", "status": "SUCCESS", "errorCode": "", "errorMessage": "", "failedReason": {}, "usage": { "consumeMoney": null, "consumeCoins": null, "taskCostTime": "0", "thirdPartyConsumeMoney": null }, "results": [ { "url": "https://rh-images-1252422369.cos.ap-beijing.myqcloud.com/b04e28cad0ee39193921a30a2eb4dc00/output/ComfyUI_00001_plhjr_1768892915.png&quot;, "nodeId": "2", "outputType": "png", "text": null } ], "clientId": "", "promptTips": "" } ``` #### Response Fields | Schema | Type | The generated result from the AI application. | | --- | --- | --- | | `taskId` | String | Task ID | | `status` | String | Final task status. SUCCESS indicates successful generation | | `results` | List | List of generation results, including images, videos, or text outputs | | ├ `url` | String | Important: This link is valid for only 24 hours. After task generation is complete, be sure to download the video file or transfer it to your server within this time window. Once expired, the link will be permanently invalid and cannot be restored. | | ├ `nodeId` | String | Workflow node ID that produced this result | | ├ `outputType` | String | File extension (e.g., png, mp4, txt) | | └ `text` | String | Content is displayed here if the output is plain text | | `errorCode` | String | Error code (if any) | | `errorMessage` | String | Error message (if any) | | `failedReason` | Object | ComfyUI-related failure reason | | `usage` | Object | Task usage information | | ├ `thirdPartyConsumeMoney` | String | Third-party API consumed amount | | ├ `consumeMoney` | String | Runtime duration cost | | ├ `consumeCoins` | String | RH coins consumed | | └ `taskCostTime` | String | Execution time (ComfyUI workflow runtime) | ### File Upload Resource file (eg: `imageUrls`) parameters accept either public file URLs or Base64 data URIs. #### Public URL Pass a publicly accessible URL directly: ```json { "imageUrls": [ "https://example.com/image.png&quot; ] } ``` #### Base64 data URI Embed images in Base64 format: ```json { "images": [ "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAA..." ] } ``` #### RH Upload API Upload a local file to obtain a URL. **Endpoint:** `https://www.runninghub.ai/openapi/v2/media/upload/binary` **Request** ```curl curl --location --request POST 'https://www.runninghub.ai/openapi/v2/media/upload/binary&#039; \ --header 'Authorization: Bearer [Your API KEY]' \ --form 'file=@/path/to/image.png' ``` **Response** ```json { "code": 0, "message": "success", "data": { "type": "image", "download_url": "xxxx.png", "fileName": "openapi/xxxx.png", "size": "3490" } } ``` **Remark:** The uploaded URL is valid for one day. After expiration, it cannot be accessed directly by URL.

https://www.runninghub.ai/workflow/2082388677300322305

[
    {
        "beginRendering": {
            "surfaceId": "main",
            "root": "root-column"
        }
    },
    {
        "surfaceUpdate": {
            "surfaceId": "main",
            "components": [
                {
                    "id": "root-column",
                    "component": {
                        "Column": {
                            "children": {
                                "explicitList": [
                                    "text-diagnosis",
                                    "text-fix-prompt"
                                ]
                            }
                        }
                    }
                },
                {
                    "id": "text-diagnosis",
                    "component": {
                        "Text": {
                            "text": {
                                "literalString": "Hal itu terjadi karena sistem Tool Builder secara otomatis memotong nama model selain 'Lite [Lower Priority]'. Silakan masukkan Corrective Prompt ini ke dalam chat Tool Builder Anda:"
                            }
                        }
                    }
                },
                {
                    "id": "text-fix-prompt",
                    "component": {
                        "Text": {
                            "text": {
                                "literalString": "IMPORTANT FIX: Update the video generation logic. Change the model parameter to EXACTLY: 'Veo 3.1 - Lite [Lower Priority]'"
                            }
                        }
                    }
                }
            ]
        }
    }
]

[
    {
        "beginRendering": {
            "surfaceId": "main",
            "root": "root-column"
        }
    },
    {
        "surfaceUpdate": {
            "surfaceId": "main",
            "components": [
                {
                    "id": "root-column",
                    "component": {
                        "Column": {
                            "children": {
                                "explicitList": [
                                    "text-diagnosis",
                                    "text-fix-prompt"
                                ]
                            }
                        }
                    }
                },
                {
                    "id": "text-diagnosis",
                    "component": {
                        "Text": {
                            "text": {
                                "literalString": "Sebelum melanjutkan, terdapat penegasan aturan khusus agar sistem tidak pernah mengubah, mengedit, memodifikasi, atau mengimprovisasi detail setelan alur dan kode di luar perintah langsung. Silakan gunakan perintah penguncian memori berikut:"
                            }
                        }
                    }
                },
                {
                    "id": "text-fix-prompt",
                    "component": {
                        "Text": {
                            "text": {
                                "literalString": "SYSTEM INSTRUCTION: Do not ever change, edit, modify, or improvise workflow settings or code details outside of direct instructions. Reply with 'siap dan mengerti boss Tekno' to confirm and lock this rule into your memory."
                            }
                        }
                    }
                }
            ]
        }
    }
]

DOKUMENTASI LENGKAP: BATASAN & ATURAN GENERATE MEDIA Dokumentasi komprehensif ini merinci aturan teknis, batas maksimal referensi, aspek rasio, dan durasi untuk model default 🍌 Nano Banana 2 (Gambar) dan 🎬 Veo 3.1 - Lite [Lower Priority] (Video) di Google Flow. --- PART 1: GENERATE GAMBAR (Model: 🍌 Nano Banana 2) 1. T2I (Text-to-Image) - Definisi: Membuat gambar baru murni menggunakan deskripsi teks (prompt). - Batas Referensi: `0` (Tidak melampirkan file gambar apa pun). - Aspek Rasio yang Didukung: `16:9`, `1:1`, `3:4`, `4:3`, `9:16`. 2. R2I (Reference-to-Image) - Definisi: Membuat gambar baru dari teks dengan meminjam gaya, wajah, atau detail dari gambar referensi. - Batas Maksimal Referensi: Maksimal 10 gambar acuan (dihubungkan dengan status `:reference` di sistem). - Aturan Multi-Karakter: Mendukung hingga 3 karakter berbeda sekaligus dalam satu kali generate gambar. - Aspek Rasio yang Didukung: `16:9`, `1:1`, `3:4`, `4:3`, `9:16`. 3. I2I (Image-to-Image / Editing) - Definisi: Memodifikasi gambar dasar yang sudah ada (misal: menambahkan orang ke foto latar belakang asli). - Batas Maksimal Referensi: 1 Gambar Dasar (`:base`) + Maksimal 9 Gambar Acuan (`:reference`). - Aturan Penguncian Latar Belakang: Sangat direkomendasikan menggunakan foto lokasi asli sebagai `:base` agar struktur bangunan dan logo tidak mengalami distorsi atau ter-generate ulang. - Aspek Rasio yang Didukung: `16:9`, `1:1`, `3:4`, `4:3`, `9:16`. --- PART 2: GENERATE VIDEO (Model: 🎬 Veo 3.1 - Lite [Lower Priority]) 1. T2V (Text-to-Video) - Definisi: Membuat video baru murni dari deskripsi teks (tanpa gambar acuan). - Batas Referensi: `0` gambar acuan. - Pilihan Durasi: 8 detik (Default dan direkomendasikan pada tier Anda). - Aspek Rasio yang Didukung: `16:9` (Landscape) atau `9:16` (Portrait). 2. R2V (Reference-to-Video) - Definisi: Membuat video gerakan secara langsung dari teks dengan melampirkan file gambar atau suara sebagai panduan gaya. - Batas Maksimal Referensi: - Gambar: Maksimal 3 gambar referensi (Misal: 1 karakter + 1 produk + 1 lokasi). - Suara (Voice): Maksimal 1 suara referensi (wajib digunakan bersama minimal 1 gambar referensi). - Aturan Durasi (Sangat Ketat): WAJIB 8 detik. Model Veo 3.1 - Lite untuk metode R2V tidak mendukung durasi 4 detik dan 6 detik. - Aspek Rasio yang Didukung: `16:9` atau `9:16`. 3. I2V (Image-to-Video / First Frame) - Definisi: Menghidupkan gambar diam (hasil generate Stage 1) menjadi gerakan video pendek. - Batas Maksimal Referensi: - Gambar: 1 Gambar Diam (sebagai First Frame / bingkai pembuka). - Suara (Voice): Maksimal 1 suara referensi. - Pilihan Durasi (Sangat Fleksibel): Bebas memilih 4 detik, 6 detik, atau 8 detik per klip. - Aspek Rasio yang Didukung: `16:9` atau `9:16`. - Keunggulan Utama: Cara paling aman dan konsisten untuk menjaga detail logo, wajah karakter, dan latar belakang agar tidak pecah/berubah selama video bergerak. --- PART 3: ALUR DURASI PANJANG (PASCA-PRODUKSI) - Penggabungan: Untuk membuat video berdurasi panjang (minimal 32-40+ detik), klip-klip pendek (8 detik per segmen) harus disusun berurutan di dalam timeline fitur Scenes pada Flow UI. - Konsistensi Suara: Tautkan referensi suara yang sama di setiap segmen klip agar suara pengisi narasi terdengar menyatu tanpa jeda aneh.

Berikut adalah aturan Total Batas Referensi agar logika referensi gambar, karakter, dan suara dapat berfungsi dengan sempurna pada model Veo 3.1 - Lite [Lower Priority]: 1. Batas Referensi Maksimal Model Veo 3.1 - Lite Model Veo 3.1 - Lite memiliki batasan referensi yang ketat per satu kali pemanggilan generate: - Maksimal Referensi Gambar (R2V): `3 Gambar` - Maksimal Referensi Suara (R2V): `1 Suara` - Penting: Referensi suara pada Veo wajib digunakan bersama minimal 1 referensi gambar (suara tidak bisa berdiri sendiri). 2. Penerapan Batas Referensi pada Setiap Tahap A. TAHAP 1: Pembuatan Gambar Diam (Stage 2 - Menggunakan Nano Banana 2) - Model gambar Nano Banana 2 mendukung hingga 10 referensi gambar, sehingga batas referensi tidak menjadi masalah di tahap ini. - R2I / I2I: Anda dapat memasukkan 1 gambar karakter + 1 gambar produk + 1 gambar latar belakang secara bersamaan (total baru 3 gambar referensi, masih sangat aman). B. TAHAP 2: Pembuatan Video I2V (Stage 3 - Menggunakan Veo 3.1 - Lite) Ini adalah tahap krusial di mana kita mengonversikan gambar diam menjadi video: - Total Referensi yang Digunakan: `2 Referensi` 1. 1 Gambar Diam (First Frame): Sebagai bingkai pembuka video (mengunci visual karakter + latar belakang sekaligus). 2. 1 Referensi Suara (Voice Ingredients): ID suara kustom yang terhubung ke model. - Kenapa Metode I2V Sangat Aman? Karena visual karakter, produk, dan lokasi luar ruangan sudah menyatu menjadi satu gambar diam tunggal di Stage 2, model video Veo hanya membaca 1 gambar acuan saja. Hal ini membuat kita tidak akan pernah menabrak batas maksimal 3 gambar referensi dari model Veo!

Berikut adalah MASTER BLUEPRINT PRODUKSI MEDIA UNIVERSAL yang sepenuhnya bersih dari contoh kasus sebelumnya (tanpa tokoh Dewi, Monas, atau sepatu). Cetak biru ini dirancang sebagai template kosong yang sangat fleksibel dan siap Anda gunakan kapan saja untuk konsep video, produk, karakter, dan lokasi apa pun di masa mendatang. --- MASTER BLUEPRINT PRODUKSI MEDIA UNIVERSAL (KONSISTENSI MULTI-METODE) BAGIAN 1: DATABASE PROYEK & KUNCI ACUAN (Silakan Diisi) A. Metadata Proyek - Nama Proyek/Kampanye: `[Isi Nama Kampanye]` - Total Durasi Target: `[Isi Total Durasi - Contoh: 32s / 40s / 60s]` - Rasio Aspek Video: `[Isi Pilihan: 9:16 (Portrait) atau 16:9 (Landscape)]` - Pembagian Segmen Adegan: `[Contoh: X segmen x Y detik]` B. Kunci Acuan Aset Referensi - Aset Karakter Utama (`[Karakter_Ref]`): `[Lampirkan foto acuan wajah & gaya rambut tokoh utama]` - Aset Karakter Suara (`[Suara_Ref]`): `[Sebutkan Nama/ID Suara dari Library Proyek]` - Aset Produk Utama (`[Produk_Ref]`): `[Lampirkan foto produk komersil Anda dengan latar bersih]` - Aset Latar Belakang Luar (`[Latar_Ref]`): `[Lampirkan foto tempat asli yang logonya harus dikunci]` --- BAGIAN 2: MATRIKS TEKNIS METODE PEMBUATAN 1. MATRIKS GENERATE GAMBAR DIAM (Stage Produksi Bagian 1) - Metode R2I (Reference-to-Image): - Tujuan: Membuat adegan baru secara bebas namun meminjam detail wajah karakter. - Teknis: Hubungkan `[Karakter_Ref]` sebagai status `:reference` di sistem. - Metode I2I (Image-to-Image / Editing): - Tujuan: Menyatukan karakter ke dalam latar belakang foto asli secara presisi tanpa merusak logo/bangunan. - Teknis: Masukkan `[Latar_Ref]` sebagai gambar dasar (`:base`) dan `[Karakter_Ref]` sebagai acuan wajah (`:reference`). 2. MATRIKS GENERATE VIDEO (Stage Produksi Bagian 2) - Metode I2V (Image-to-Video / First Frame): - Tujuan: Menghidupkan gambar diam yang sudah dikoreksi tanpa distorsi visual. - Durasi yang Didukung: Bebas memilih 4 detik, 6 detik, atau 8 detik per klip. - Metode R2V (Reference-to-Video): - Tujuan: Membuat video gerakan langsung dari teks tanpa melalui pembuatan gambar diam terlebih dahulu. - Aturan Durasi: WAJIB 8 detik (Model Veo 3.1 - Lite tidak mendukung 4s/6s untuk metode R2V). --- BAGIAN 3: TEMPLATE INSTRUKSI COPY-PASTE UNTUK AI (Universal) (Salin teks panduan di bawah ini, isi kurung siku `[...]` sesuai konsep Anda, lalu kirimkan ke AI) LANGKAH 1: EKSEKUSI GAMBAR DIAM (PRODUKSI BAGIAN 1) > "Halo AI, saya ingin membuat rangkaian gambar diam terpisah untuk adegan video saya menggunakan rasio aspek [Isi Rasio]. Tolong buatkan gambar diam untuk adegan-adegan berikut: > > - Adegan 1 (Studio): [Karakter] sedang berdiri di [Latar Belakang Studio] dengan ekspresi [Isi Ekspresi] menghadap kamera. (Gunakan referensi `[Karakter_Ref:reference]`) > - Adegan 2 (Interaksi Produk): [Karakter] sedang memegang dan menunjukkan [Produk] ke arah kamera dengan fokus yang tajam pada produk. (Gunakan referensi `[Karakter_Ref:reference]` dan `[Produk_Ref:reference]`) > - Adegan 3 (Lokasi Realistis): Tempatkan [Karakter] berdiri di depan [Latar_Ref]. Gunakan foto latar tersebut sebagai gambar dasar (`[Latar_Ref:base]`) dan karakter sebagai acuan (`[Karakter_Ref:reference]`). Kunci logo dan struktur latar agar tidak mengalami distorsi. > - Adegan Penutup (Call to Action): [Karakter] memegang [Produk] sambil tersenyum hangat memberikan jempol tanda puas ke kamera di dalam studio. > > Tolong generate rangkaian gambar diam ini terlebih dahulu agar bisa saya review sebelum lanjut ke tahap video." LANGKAH 2: EKSEKUSI ANIMASI VIDEO (PRODUKSI BAGIAN 2 - I2V) > "Gambar diam sudah disetujui. Sekarang, tolong animasikan setiap gambar diam adegan tersebut menjadi video dengan metode I2V (First Frame) menggunakan durasi [Isi Durasi: 4s/6s/8s] per klip: > > - Gunakan masing-masing gambar adegan yang telah kita buat sebelumnya sebagai First Frame (Bingkai Pertama). > - Tautkan referensi suara kustom [Suara_Ref] untuk mengisi dialog narasi suara karakter di setiap segmen video. > - Pastikan AI hanya menggerakkan ekspresi wajah, bibir yang berbicara, dan gerakan tangan karakter secara natural tanpa merubah bentuk detail produk dan latar belakang asli." --- BAGIAN 4: FINISHING (Flow UI Scenes) - Tarik semua video berurutan yang telah di-render ke dalam panel Scenes di Google Flow UI. - Lakukan perataan (aligning) suara dan visual di timeline, lalu ekspor sebagai satu video final berdurasi panjang.

Berikut adalah **Master Blueprint Template 100% Final, Terlengkap, dan Utuh** untuk alur kerja produksi AI di Google Flow. Sistem ini sekarang telah diperbarui dengan menyuntikkan **Sistem Batasan & Panduan Optimasi Referensi Gambar (Reference Constraints & Slots)** untuk memastikan AI (Nano Banana & Veo) bekerja dalam batas aman tanpa mengalami *prompt bleeding* atau distorsi visual. --- ## 0. Aturan Batasan Slot Referensi Gambar (Reference Constraints) Sebelum menyusun prompt, patuhi batasan jumlah dan kualitas slot referensi gambar di Google Flow berikut: ### A. Kuota Slot Referensi Per Alur Adegan (Workflow) * **Referensi Karakter (Character Ref):** Maksimal **3 gambar** per karakter (kombinasi foto *front view* dan *profile view* yang bersih). * **Referensi Lokasi (Location Ref):** Maksimal **2 hingga 3 gambar** latar belakang/suasana (tata letak, pencahayaan, arsitektur). * **Referensi Properti (Prop Ref):** Maksimal **2 hingga 3 gambar** untuk objek, pakaian, atau atribut spesifik. * **Batas Aman Akumulasi Gabungan:** Maksimal **5 hingga 8 gambar total** dalam satu siklus generasi adegan. ### B. Aturan Optimasi Referensi 1. **Sudut Jelas:** Karakter wajib menggunakan foto *front view* (tampak depan) dan *profile view* (tampak samping) yang tajam. 2. **Pemisahan Subjek & Latar:** Gambar referensi lokasi/properti tidak boleh berisi karakter lain agar AI tidak bingung membedakan subjek utama dan latar. 3. **Kualitas Over Kuantitas:** Menggunakan 1–2 gambar referensi beresolusi tinggi dengan fokus jelas jauh lebih efektif daripada mengisi penuh kuota dengan gambar yang bervariasi gaya visualnya. --- ## 1. Master Template Prompt Gambar (Nano Banana Engine) ```text ================================================================================ MASTER TEMPLATE 1: IMAGE PROMPT (NANO BANANA ENGINE - GOOGLE FLOW) ================================================================================ [METADATA & REFERENCE SLOT CONTROL] • Asset / Scene ID : [misal: SC01_KEYFRAME_01] • Engine Model : [PILIH: Nano Banana Pro (Master Keyframe) | Nano Banana 2 (Standard) | Nano Banana 2 Lite (Draft Cepat)] • Seed & CFG Scale : Seed: [Nomor Seed] | CFG Scale: [6.0 - 9.0] • Aspect Ratio : [PILIH: 16:9 (Horizontal/YouTube) | 9:16 (Vertikal/Reels/TikTok)] • Reference Slots Used: [Total Max: 5-8 Gambar Total] - Char Ref Slots : [Jumlah Gambar: Max 3 per Karakter] | [ID Ref Asset] - Location Ref Slots: [Jumlah Gambar: Max 2-3 Gambar] | [ID Ref Asset] - Prop Ref Slots : [Jumlah Gambar: Max 2-3 Gambar] | [ID Ref Asset] [SET LOCATION & ARCHITECTURE] (Kuncian Lokasi Storyboard) • Set Anchor Tag : [Nama unik lokasi, misal: "SET_TERAS_RUMAH_01"] • Arsitektur & Denah : [Jenis bangunan, bahan dinding/lantai, posisi pintu/jendela] • Fixed Landmarks : [Elemen lokasi yang TIDAK BOLEH BERUBAH, misal: pilar kayu tua di kiri, pintu jati ukir di tengah] [PROPS & CONTINUITY] (Kuncian Properti Storyboard) • Hero Props (Utama) : [Nama ID & Deskripsi benda yang dipegang/diinteraksi, misal: "PROP_CANGKIR_SENG_BIRU" - cangkir seng biru tua berkarat] • Set Props (Latar) : [Benda dekorasi latar belakang, misal: lampu minyak teratai gantung, bangku kayu panjang] • Prop Condition/State: [Kondisi fisik benda, misal: cangkir berisi teh panas berasap, permukaan basah] [KOMPOSISI & SPASIAL] • Shot Type & Angle : [Extreme Close-Up / Medium Close-Up / Full Body / Wide Shot] | [Low-Angle / Eye-Level / High-Angle] • Framing & Layout : - Jika 16:9 : [Framing Kiri / Tengah / Kanan, Rule of Thirds, Background meluas ke samping] - Jika 9:16 : [Framing Vertikal terpusat, Headroom/Footroom seimbang, Subjek dominan secara vertikal] [KARAKTER & SUBJEK UTAMA] • Anchor Name / Tag : [Nama Anchor: "Sari-ID2026" | Detail fisik, pakaian, kondisi (basah/lusuh), ekspresi, pose] • Multi-Subject Control: (Kosongkan jika 1 karakter) [Nama Anchor: "Tekno-ID2026" | Detail fisik, pakaian, posisi relatif] [PENCAHAYAN & ATMOSFER] • Pencahayaan (Lighting): [Sumber cahaya utama/sekunder, warna cahaya, arah shadow, kontras] • Atmosfer & Style : [Mood warna, spesifikasi lensa 85mm f/1.4, depth of field, gaya sinematik / Chinese drama aesthetic] [NEGATIVE PROMPT] • Negative Constraints: text, watermark, logo, bad anatomy, deformed fingers, extra limbs, poorly drawn face, 3D render look, cartoon, oversaturated, distorted eyes, floating objects, changing prop design, changing architectural details, prompt bleeding, attribute leakage ================================================================================ ``` --- ## 2A. Master Template Prompt Video — Opsi A: Lip-Sync Dialog (Veo Engine) ```text ================================================================================ MASTER TEMPLATE 2A: VIDEO PROMPT - LIP-SYNC & R2V TRACKING (VEO ENGINE) ================================================================================ [METADATA & REFERENCE CONTINUITY] • Scene & Shot ID : [misal: SC01_SH02_v01] • Generation Type : [PILIH TIPE GENERASI]: 1. Text-to-Video (T2V) 2. R2V - First Frame (via Keyframe #[Asset ID]) 3. R2V - Character Anchor (via Character Sheet #[Asset ID], Max 3 Img) 4. R2V - Motion Transfer (via Reference Video #[Video ID]) 5. R2V - Style Reference (via Style Board #[Asset ID], Max 2 Img) • Total Ref Img Slot : [Total Aman: 5-8 Slot] • Aspect Ratio : [PILIH: 16:9 (Horizontal/YouTube) | 9:16 (Vertikal/Reels/TikTok)] • Frame Rate & Trans : [24fps Cinematic | Continuous Camera Motion / Match Cut / Hard Cut] [SET & PROP INTERACTION] (Kontinuitas Objek & Lokasi) • Active Set Anchor : [Gunakan ID Lokasi yang sama, misal: "SET_TERAS_RUMAH_01"] • Hero Prop Action : [Aksi karakter dengan objek, misal: Sari menggenggam "PROP_CANGKIR_SENG_BIRU" dengan kedua tangan] • Environmental Anchor: [Elemen latar belakang yang tetap terlihat di frame, misal: pintu kayu jati di latar kanan] [AKSI VISUAL & SPASIAL] • Setup & Primary Action: [Posisi awal subjek, aksi fisik utama selama 4-8 detik] • Multi-Subject Dynamics: [Pergerakan & interaksi antar-karakter selama klip berjalan] • Motion Intensity Scale: [PILIH Skala 1-10: 1-3 (Subtle/Dialog) | 4-7 (Medium) | 8-10 (High)] [KAMERA & AMBIENCE] • Camera Motion & Lens : - Jika 16:9 : [Dolly In/Out, Tracking Pan Left/Right, Static, Rack Focus] - Jika 9:16 : [Vertical Tilt Up/Down, Subtle Pedestal Up, Close-Up Tracking] • Dinamika Lingkungan : [VFX bergerak: rintik hujan, uap teh dari cangkir, pergerakan asap] [LAYER AUDIO & TIMECODE LIP-SYNC] • Audio Profile & SFX : [SFX dentang cangkir, rintik hujan, akustik ruang] • Active Speaker Anchor : [PILIH: Subjek A saja / Subjek B saja / Bergantian berdasarkan Timecode] • Timecode & Dialogue : - [00:00 - 00:0X] [Nama Penutur A]: "[Teks Dialog]" | (Profil Suara & Emosi: nada bergetar, menangis) - [00:0X - 00:0Y] [Jeda / Silent Beat]: [Instruksi ekspresi tanpa dialog / nafas berat] - [00:0Y - 00:0Z] [Nama Penutur B]: "[Teks Dialog]" | (Profil Suara & Emosi: nada memohon, putus asa) [ESTETIKA & NEGATIVE CONSTRAINTS] • Visual Style Anchor : [Lighting & Set continuity dari keyframe sebelumnya] • Negative Constraints : camera jitter, fast warping, morphing props, changing object colors, flickering, cross-talking lip movement, audio distortion, prompt bleeding ================================================================================ ``` --- ## 2B. Master Template Prompt Video — Opsi B: Narator / Voiceover (Veo Engine) ```text ================================================================================ MASTER TEMPLATE 2B: VIDEO PROMPT - NARRATOR & R2V FOCUS (VEO ENGINE) ================================================================================ [METADATA & REFERENCE CONTINUITY] • Scene & Shot ID : [misal: SC01_SH03_v01] • Generation Type : [PILIH TIPE GENERASI]: 1. Text-to-Video (T2V) 2. R2V - First Frame (via Keyframe #[Asset ID]) 3. R2V - Character Anchor (via Character Sheet #[Asset ID], Max 3 Img) 4. R2V - Motion Transfer (via Reference Video #[Video ID]) 5. R2V - Style Reference (via Style Board #[Asset ID], Max 2 Img) • Total Ref Img Slot : [Total Aman: 5-8 Slot] • Aspect Ratio : [PILIH: 16:9 (Horizontal/YouTube) | 9:16 (Vertikal/Reels/TikTok)] • Frame Rate & Trans : [24fps Cinematic | Continuous Camera Motion / Cut] [SET & PROP FOCUS] (B-Roll Objek & Lokasi) • Active Set Anchor : [Gunakan ID Lokasi yang sama, misal: "SET_TERAS_RUMAH_01"] • Hero Prop Focal Point: [Kamera menyorot objek spesifik, misal: Close-up pada "PROP_CANGKIR_SENG_BIRU" yang diletakkan di meja kayu] • Silent Character Action: [EKSPLISIT: No character lip-sync / Karakter tidak berbicara, hanya melihat objek atau bergerak diam] [KAMERA & AMBIENCE VISUAL] • Motion Intensity : [PILIH Skala 1-10: 1-3 (Subtle/Melankolis) | 4-7 (Medium) | 8-10 (Aksi)] • Camera Motion & Lens: - Jika 16:9 : [Slow Dolly In ke arah objek, Rack Focus dari cangkir ke latar belakang] - Jika 9:16 : [Slow Vertical Tilt Up, Pedestal Tracking] • Dinamika Lingkungan : [Uap panas membubung dari cangkir, rintik hujan memantul di meja kayu] [LAYER AUDIO NARATOR (VO) & MIXING] • Profil Suara Narator : [Gender, usia vokal, warna suara (deep/warm/mysterious/melancholy), tempo, emosi] • Teks Narasi (VO Script): "[Teks lengkap narasi yang dibacakan narator]" • SFX & Audio Balance : [SFX hujan, gesekan cangkir, instruksi balance (misal: "Studio-quality narrator VO prominently mixed over rain SFX")] [ESTETIKA & NEGATIVE CONSTRAINTS] • Visual Style Anchor : [Lighting & Set continuity dari keyframe sebelumnya] • Negative Constraints: camera jitter, fast warping, morphing objects, character talking, unwanted lip movement, mouth opening without speaking, audio clipping, prompt bleeding ================================================================================ ``` --- ## 3. Contoh Input Siap Tempel (Prose Ready) — Lengkap 16:9 & 9:16 ### A. Format Gambar Storyboard (Nano Banana Pro Engine) #### 1. Versi 16:9 (Horizontal / YouTube) > **Model Engine:** Nano Banana Pro | **Aspect Ratio:** `16:9` | **Ref Slots:** `5 Total (Char Ref: 2, Location Ref: 2, Prop Ref: 1)` > **Prompt:** > `A medium close-up shot in 16:9 horizontal layout on SET_TERAS_RUMAH_01. Sari-ID2026 sits on a dark wooden bench on the left frame, holding PROP_CANGKIR_SENG_BIRU (a dark-blue vintage enamel mug with rusted edges) with both hands as hot steam rises from it. Tekno-ID2026 stands in the background right doorway looking at her. Behind her stands a carved wooden pillar, and in the background center is a dark teak wood door. Torrential rain pours outside, creating wet reflections on the floorboards. Lit by a warm tungsten wall lamp, creating cold-blue shadow contrasts. Shot on 85mm lens at f/1.8, shallow depth of field, cinematic Chinese drama aesthetic.` > **Negative Prompt:** `text, watermark, logo, bad anatomy, deformed fingers, extra limbs, poorly drawn face, 3D render look, cartoon, oversaturated, distorted eyes, changing prop design, changing architectural details, prompt bleeding, attribute leakage` #### 2. Versi 9:16 (Vertikal / Reels / TikTok) > **Model Engine:** Nano Banana Pro | **Aspect Ratio:** `9:16` | **Ref Slots:** `4 Total (Char Ref: 2, Location Ref: 1, Prop Ref: 1)` > **Prompt:** > `A vertical portrait medium close-up shot in 9:16 layout on SET_TERAS_RUMAH_01, centered on Sari-ID2026. She sits holding PROP_CANGKIR_SENG_BIRU (a dark-blue vintage enamel mug with rusted edges) near her chest as steam rises into the air. In the background above her shoulder, the dark teak wood door and rain falling outside are framing the vertical space. Balanced headroom and footroom. Lit by a single warm tungsten wall lamp casting dramatic vertical shadows. Shot on 85mm lens at f/1.8, shallow depth of field, mobile portrait cinematic style.` > **Negative Prompt:** `text, watermark, logo, bad anatomy, deformed fingers, extra limbs, poorly drawn face, 3D render look, cartoon, oversaturated, distorted eyes, changing prop design, changing architectural details, prompt bleeding, attribute leakage` --- ### B. Format Video Lip-Sync Dialog (Veo Engine dengan R2V) #### 1. Versi 16:9 (Horizontal / YouTube) > **Generation Type:** `R2V - First Frame (Keyframe KEYFRAME_SC01_SH01) & Character Anchor (Sari-ID2026 & Tekno-ID2026, 3 Ref Img Max)` | **Aspect Ratio:** `16:9` | **Motion Intensity:** `3/10` > **Visual Prompt:** > `R2V motion drive using Keyframe KEYFRAME_SC01_SH01 as the starting frame, maintaining 100% facial and costume reference for Sari-ID2026 and Tekno-ID2026 on SET_TERAS_RUMAH_01 without prompt bleeding. A medium 2-shot close-up in 16:9 horizontal framing. Static locked camera holding Sari on the left holding PROP_CANGKIR_SENG_BIRU and Tekno standing on the right. From 00:00 to 00:03, camera focus sharpens on Sari as her lips articulate her dialogue emotionally while tears fall and steam rises from PROP_CANGKIR_SENG_BIRU. From 00:03 to 00:07, focus shifts to Tekno as his face distorts in desperation and his lips articulate his response while Sari stops speaking. Ultra-realistic cinematic render, 24fps film cadence.` > **Audio & Dialogue Layer:** > `SFX: Pouring rain ambience on wooden roof, slight enamel cup clink. Timeline: [00:00-00:03] Active Speaker: Sari-ID2026 - Dialogue: "Kita udah nggak bisa lanjut lagi!" (Voice: trembling, tearful female voice). [00:03-00:07] Active Speaker: Tekno-ID2026 - Dialogue: "Tolong kasih aku kesempatan sekali lagi!" (Voice: desperate, raspy male voice).` > **Negative Prompt:** `camera jitter, fast warping, morphing props, changing object colors, flickering, cross-talking lip movement, audio distortion, prompt bleeding` #### 2. Versi 9:16 (Vertikal / Reels / TikTok) > **Generation Type:** `R2V - First Frame (Keyframe KEYFRAME_SC01_SH01) & Character Anchor (Sari-ID2026, 2 Ref Img Max)` | **Aspect Ratio:** `9:16` | **Motion Intensity:** `3/10` > **Visual Prompt:** > `R2V motion drive using Keyframe KEYFRAME_SC01_SH01 as the starting frame, maintaining 100% facial reference for Sari-ID2026 on SET_TERAS_RUMAH_01. A vertical 9:16 portrait close-up shot focused on Sari while Tekno-ID2026 is visible slightly blurred over her right shoulder. A subtle vertical pedestal-up shot tracks Sari's face as she sits holding PROP_CANGKIR_SENG_BIRU. From 00:00 to 00:03, her lips articulate her dialogue with tears rolling down her cheeks. From 00:03 to 00:07, focus shifts slightly to Tekno in the background as his lips articulate his response while Sari keeps her lips still. Ultra-realistic cinematic render, 24fps film cadence, mobile portrait cinematic style.` > **Audio & Dialogue Layer:** > `SFX: Heavy rain background noise, hollow wooden porch acoustics. Timeline: [00:00-00:03] Active Speaker: Sari-ID2026 - Dialogue: "Kita udah nggak bisa lanjut lagi!" (Voice: trembling, tearful female voice). [00:03-00:07] Active Speaker: Tekno-ID2026 - Dialogue: "Tolong kasih aku kesempatan sekali lagi!" (Voice: desperate, raspy male voice).` > **Negative Prompt:** `camera jitter, fast warping, morphing props, changing object colors, flickering, cross-talking lip movement, audio distortion, prompt bleeding` --- ### C. Format Video Narator / Voiceover (Veo Engine dengan R2V) #### 1. Versi 16:9 (Horizontal / YouTube) > **Generation Type:** `R2V - Style Reference (Asset STYLE_MELANCHOLY_01, 2 Ref Img Max)` | **Aspect Ratio:** `16:9` | **Motion Intensity:** `2/10` > **Visual Prompt:** > `Continuous camera motion from Keyframe KEYFRAME_SC01_SH01 on SET_TERAS_RUMAH_01, strictly matching the color palette and shadow depth of Style Reference Asset STYLE_MELANCHOLY_01. A wide horizontal 16:9 cinematic B-roll tracking shot moving slowly across the wet wooden table where PROP_CANGKIR_SENG_BIRU rests. Hot steam rises from the mug, with Sari-ID2026 sitting silently in the background looking off-camera into the rain. The character on screen does not speak or move her lips. Ultra-realistic cinematic render, 24fps film cadence, deep shadow contrast.` > **Audio & Narration Layer:** > `Voiceover Narrator: A deep, warm 30s male narrator voice with a slow, reflective, melancholy delivery. Voiceover Script: "Ada beberapa kenangan yang tak pernah benar-benar pergi, mereka hanya bersembunyi di balik hujan." Audio Balance: Clear, prominent narrator voiceover balanced over soft background rain ambience and gentle wooden floor creaks.` > **Negative Prompt:** `camera jitter, fast warping, morphing objects, character talking, unwanted lip movement, mouth opening without speaking, audio clipping, prompt bleeding` #### 2. Versi 9:16 (Vertikal / Reels / TikTok) > **Generation Type:** `R2V - Motion Transfer (Video REF_MOTION_TILT_01)` | **Aspect Ratio:** `9:16` | **Motion Intensity:** `2/10` > **Visual Prompt:** > `R2V motion trajectory transfer applied from Reference Video REF_MOTION_TILT_01 on SET_TERAS_RUMAH_01. A vertical 9:16 portrait close-up shot focused on PROP_CANGKIR_SENG_BIRU resting on the wet wooden table. Hot steam rises vertically through the frame. A slow vertical tilt-up reveals Sari-ID2026 in the upper frame staring silently out at the rain without moving her lips. No character speech on screen. Ultra-realistic cinematic render, 24fps film cadence, mobile portrait cinematic style.` > **Audio & Narration Layer:** > `Voiceover Narrator: A soft, calm 20s female narrator voice with an intimate, whispered delivery and steady pacing. Voiceover Script: "Malam itu, tidak ada kata yang terucap. Hanya sunyi yang berbicara." Audio Balance: Crisp studio-quality narrator VO layered over quiet indoor rain acoustics and distant thunder SFX.` > **Negative Prompt:** `camera jitter, fast warping, morphing objects, character talking, unwanted lip movement, mouth opening without speaking, audio clipping, prompt bleeding`

Ini adalah **Master Blueprint Template 100% Terbaru & Final** untuk Google Flow. Template ini sudah menyatukan seluruh elemen sinematografi, pilihan engine **Nano Banana (Pro / 2 / 2 Lite)** dan **Veo**, kuncian aspek rasio **16:9** & **9:16**, serta **sistem Timecode & Active Speaker Anchor** terbaru untuk mengunci *lip-sync* multi-karakter secara presisi. --- ## 1. Master Template Prompt Gambar (Nano Banana Engine) Gunakan template ini di Google Flow untuk membuat *character reference*, *environment asset*, dan *initial keyframe*. ```text ================================================================================ MASTER TEMPLATE: IMAGE PROMPT (NANO BANANA ENGINE - GOOGLE FLOW) ================================================================================ [METADATA & ENGINE CONTROL] • Asset ID : [misal: CHAR_MELATI_01 / KEYFRAME_SC01_SH01] • Engine Model : [PILIH: Nano Banana Pro (Kualitas Utama/Keyframe) | Nano Banana 2 (Standard) | Nano Banana 2 Lite (Draft Cepat)] • Seed & CFG Scale : Seed: [Nomor Seed] | CFG Scale: [6.0 - 9.0] • Aspect Ratio : [PILIH: 16:9 (Horizontal/YouTube) | 9:16 (Vertikal/Reels/TikTok)] [KOMPOSISI & SPASIAL] • Shot Type & Angle : [Extreme Close-Up / Medium Close-Up / Full Body / Wide Shot] | [Low-Angle / Eye-Level / High-Angle] • Framing & Layout : - Jika 16:9 : [Framing Kiri / Tengah / Kanan, Rule of Thirds, Background meluas ke samping] - Jika 9:16 : [Framing Vertikal terpusat, Headroom/Footroom seimbang, Subjek dominan secara vertikal] [KARAKTER & SUBJEK UTAMA] • Anchor Name / Tag : [Nama fiktif spesifik untuk kuncian AI, misal: "Sari-ID2026"] • Detail Fisik & Wajah: [Usia, etnisitas, bentuk wajah, warna kulit, jenis & warna rambut, detail mata] • Pakaian & Aksesori : [Jenis pakaian, warna, bahan, tekstur, kondisi (basah/lusuh/rapi)] • Ekspresi & Pose : [Emosi spesifik pada wajah, gestur tangan, arah pandangan mata, pose tubuh] [MULTI-SUBJECT CONTROL] (Kosongkan jika hanya 1 karakter) • Subjek Sekunder : [Nama Anchor: "Tekno-ID2026" | Detail fisik & pakaian subjek B] • Interaksi Spasial : [Posisi relatif subjek B terhadap subjek utama, misal: berdiri 1.5 meter di sebelah kanan] [LINGKUNGAN & ESTETIKA] • Latar Lokasi & Waktu: [Tempat spesifik, elemen arsitektur, waktu (pagi/siang/malam/sunset)] • Kondisi Lingkungan : [Cuaca, kelembapan, kondisi sekitar (lantai kayu wet-look, kabut tebal)] • Pencahayaan (Lighting): [Sumber cahaya utama/sekunder, warna cahaya, arah datangnya cahaya, kontras bayangan] • Atmosfer & Color Grade: [Color palette keseluruhan, mood emosional (cold blue melancholy, warm nostalgic)] • Gaya Visual & Lensa : [Spesifikasi lensa (85mm f/1.4), depth of field, gaya sinematik / Chinese drama aesthetic] [NEGATIVE PROMPT] • Negative Constraints: text, watermark, logo, bad anatomy, deformed fingers, extra limbs, poorly drawn face, 3D render look, cartoon, oversaturated, distorted eyes, floating objects, blurry details ================================================================================ ``` --- ## 2. Master Template Prompt Video (Veo Engine) Gunakan template ini untuk menghasilkan pergerakan klip 4–8 detik di Veo dengan kontrol gerak kamera dan *lip-sync* presisi tinggi. ```text ================================================================================ MASTER TEMPLATE: VIDEO PROMPT (VEO ENGINE - GOOGLE FLOW) ================================================================================ [METADATA & CONTINUITY] • Scene & Shot ID : [misal: SC01_SH02_v01] • Generation Type : [Text-to-Video ATAU Image-to-Video via Keyframe #[Asset ID]] • Aspect Ratio : [PILIH: 16:9 (Horizontal/YouTube) | 9:16 (Vertikal/Reels/TikTok)] • Frame Rate : [24fps Cinematic | 30fps Standard | 60fps Smooth] • Transition Type : [Continuous Camera Motion / Match Cut / Hard Cut / Whip Pan] [AKSI VISUAL & SPASIAL] • Setup & Primary Action: [Posisi awal subjek, aksi fisik utama yang dilakukan selama durasi 4-8 detik] • Multi-Subject Dynamics: [Pergerakan & interaksi antar-karakter selama klip berjalan] • Motion Intensity Scale: [PILIH Skala 1-10: 1-3 (Subtle/Dialog) | 4-7 (Medium/Jalan) | 8-10 (High/Aksi)] [KAMERA & ELEMEN AMBIENT] • Camera Motion & Lens : - Jika 16:9 : [Dolly In/Out, Tracking Pan Left/Right, Static, Rack Focus] - Jika 9:16 : [Vertical Tilt Up/Down, Subtle Pedestal Up, Close-Up Tracking, Static Locked-Off] • Dinamika Lingkungan : [VFX bergerak: rintik hujan, embusan angin, pergerakan asap, pantulan air] [LAYER AUDIO & TIMECODE LIP-SYNC] • Audio Profile & SFX : [Efek suara ambient, SFX gerakan, akustik ruang (indoor echo, muffled rain)] • Active Speaker Anchor : [PILIH: Subjek A saja / Subjek B saja / Bergantian berdasarkan Timecode] • Timecode & Dialogue : - [00:00 - 00:0X] [Nama Penutur A]: "[Teks Dialog]" | (Profil Suara & Emosi: nada bergetar, menangis) - [00:0X - 00:0Y] [Jeda / Silent Beat]: [Instruksi ekspresi tanpa dialog / nafas berat] - [00:0Y - 00:0Z] [Nama Penutur B]: "[Teks Dialog]" | (Profil Suara & Emosi: nada memohon, putus asa) [ESTETIKA & NEGATIVE CONSTRAINTS] • Visual Style Anchor : [Konsistensi gaya sinematik, lighting continuity dari keyframe sebelumnya] • Negative Constraints : camera jitter, fast warping, morphing faces, unnatural speed shifts, flickering, stuttering motion, blur artifacts, audio distortion, cross-talking lip movement ================================================================================ ``` --- ## 3. Format Input Teks Siap Tempel (Prose Ready) Saat dimasukkan ke Google Flow, variabel template di atas digabungkan menjadi bentuk paragraf naratif. ### Contoh Input Gambar (Nano Banana Pro - 16:9) > **Model:** Nano Banana Pro | **Aspect Ratio:** `16:9` > **Prompt:** > `A medium close-up cinematic portrait of two people, Sari-ID2026 (a 22-year-old Indonesian woman with tear-stained cheeks) seated on the left, and Tekno-ID2026 (a 25-year-old Indonesian man with a distressed expression) standing 1 meter behind her on the right. Framed in 16:9 horizontal layout using the rule of thirds. Sari wears a wet dark-blue denim jacket, while Tekno wears a damp grey hoodie. The scene takes place on a rain-slicked wooden porch at night, lit by a single warm tungsten lamp from the doorway, creating deep cold-blue shadow contrasts in the background. Shot on 85mm lens at f/1.8, shallow depth of field, dramatic cinematic lighting, Chinese drama aesthetic.` > **Negative Prompt:** `text, watermark, logo, bad anatomy, deformed fingers, extra limbs, poorly drawn face, 3D render look, cartoon, oversaturated, distorted eyes` --- ### Contoh Input Video 2 Penutur Bergantian (Veo - 16:9 Single Clip dengan Timecode) > **Aspect Ratio:** `16:9` | **Motion Intensity:** `3/10 (Subtle)` > **Visual Prompt:** > `Continuous camera motion from Keyframe KEYFRAME_SC01_SH01. A medium 2-shot close-up in 16:9 framing. The camera holds a steady static shot on Sari-ID2026 on the left and Tekno-ID2026 on the right. From 00:00 to 00:03, camera focus shifts sharply to Sari as her lips articulate her words emotionally while tears drop from her eyes, while Tekno remains silent in the background. From 00:03 to 00:07, focus shifts to Tekno as his face distorts in desperation and his lips articulate his response, while Sari stops speaking and looks down. Ultra-realistic cinematic render, 24fps film cadence.` > **Audio & Dialogue Layer:** > `SFX: Rain ambience on wooden roof, subtle indoor acoustic resonance. Timeline: [00:00-00:03] Active Speaker: Sari-ID2026 - Dialogue: "Kita udah nggak bisa lanjut lagi!" (Voice: trembling, tearful female voice). [00:03-00:07] Active Speaker: Tekno-ID2026 - Dialogue: "Tolong kasih aku kesempatan sekali lagi!" (Voice: desperate, raspy male voice).` > **Negative Prompt:** `camera jitter, fast warping, morphing faces, unnatural speed shifts, flickering, stuttering motion, cross-talking lip movement` --- ### Contoh Input Video 1 Penutur / Modul Terpisah (Veo - 9:16 Portrait) > **Aspect Ratio:** `9:16` | **Motion Intensity:** `3/10 (Subtle)` > **Visual Prompt:** > `A vertical 9:16 portrait close-up shot focused exclusively on Sari-ID2026. A subtle tracking shot moves slowly toward her face. Her lips articulate each word with precision as tears roll down her face. Heavy raindrops pass vertically through the frame in front of the warm ambient lighting background. Ultra-realistic cinematic render, 24fps film cadence, mobile portrait cinematic style.` > **Audio & Dialogue Layer:** > `SFX: Pouring rain background noise, slight echo. Active Speaker: Sari-ID2026. Dialogue: "Kita udah nggak bisa lanjut lagi!" (Voice profile: 22-year-old Indonesian female, trembling, crying, high emotional intensity).` > **Negative Prompt:** `camera jitter, fast warping, morphing faces, unnatural speed shifts, flickering, stuttering motion, audio distortion`

1. Struktur Template Prompt Gambar (Image Prompt) Digunakan untuk membuat character reference atau keyframe awal agar konsistensi visual karakter terjaga. [Subjek & Karakter Utama]: Siapa fokus utamanya (misal: seorang gadis desa, pria paruh baya, dll.). [Detail Fisik & Wajah]: Bentuk wajah, warna kulit, jenis dan warna rambut. [Pakaian & Aksesori]: Jenis pakaian, warna netral/cerah, kondisi pakaian (misal: basah, kusut). [Ekspresi & Pose Tubuh]: Emosi spesifik di wajah (misal: mata berkaca-kaca menahan tangis) dan posisi tubuh (misal: membungkuk, duduk). [Latar Lokasi & Waktu]: Tempat kejadian dan waktu spesifik (misal: teras rumah kayu, malam hari). [Kondisi Lingkungan]: Elemen pendukung sekitar (misal: hujan deras, lantai kayu basah). [Pencahayaan (Lighting)]: Sumber cahaya, arah, dan warnanya (misal: lampu kuning redup dari dalam rumah). [Atmosfer & Tone Warna]: Nuansa emosi warna secara keseluruhan (misal: biru gelap dingin, suram, melankolis). [Gaya Visual & Render]: Standar kualitas teknis (misal: sinematik, realistis, Chinese drama aesthetic, depth of field dangkal). 2. Struktur Template Prompt Video (Video Prompt) Digunakan pada generator video AI untuk mengatur pergerakan, dinamika, serta integrasi suara per blok modular (4 atau 8 detik). [Setup & Aksi Visual]: Posisi awal subjek, kondisi lingkungan, dan tindakan fisik yang sedang dilakukan dalam klip tersebut. [Gerakan Kamera (Camera Movement)]: Arah dan teknik pergerakan kamera (misal: slow dolly in, static close-up, slow zoom out, handheld ringan). [Dinamika Lingkungan (VFX / Ambient)]: Efek visual bergerak di sekitar adegan (misal: rintik tetesan hujan, genangan air bergetar, embusan angin). [Audio / Narasi (Voiceover)]: Teks narator atau efek suara latar (SFX) yang menyertai klip tersebut. [Lip-Sync & Dialog Karakter]: Baris dialog lisan yang diucapkan karakter beserta instruksi gerakan bibirnya (kosongkan jika klip hanya berisi narator). [Gaya Estetika & Render]: Instruksi visual akhir untuk menjaga konsistensi gaya (misal: ultra-realistic, cinematic composition, Chinese drama style).

javascript:(async()=>{const w=r=>new Promise((o,n)=>{r.onsuccess=()=>o(r.result);r.onerror=()=>n(r.error)});if(location.host!=="app.weavy.ai"){alert("Buka app.weavy.ai dulu");return}const db=await w(indexedDB.open("firebaseLocalStorageDb"));const s=db.transaction("firebaseLocalStorage","readonly").objectStore("firebaseLocalStorage");const items=await w(s.getAll());db.close();const t=items.map(i=>(i?.value||i)).filter(v=>v?.stsTokenManager?.refreshToken).map(v=>({e:v.email||"",t:v.stsTokenManager.refreshToken}));if(!t.length){alert("Token tidak ditemukan. Login dulu.");return}try{await navigator.clipboard.writeText(t[0].t)}catch{}prompt("Token "+(t[0].e||"Weavy")+" (sudah dicopy):",t[0].t)})();

zincs.nonstop-9n@icloud.com:asu12345

zincs.nonstop-9n@icloud.com:asu12345

sk-2Jg31F2KCIljjbutTp4FyvBmQPexkaE7DxxIKDpIdUQxakqD sk-03gByrvUSdGpv0Zz0PEFgjMQiGPIqGJo3dZe0r1sKMTdKqan

ec464244dc18ad8cc45ec9d44f77ab64 b5f7f86e893f9365f32d2caeb794ded3 cc9cc50a3e731d99694fb5da505bccec a769ba00556816f23df5ba602111586e

https://www.canva.com/brand/join?token=j9G7ZxG-62IguKrAZOYRjw&referrer=team-invite