Category | Feature | Specifications/Limit |
Basic image processing | Common limits | Supported format: JPG, BMP, GIF, PNG, WebP, as well as HEIF decoding and processing Size limit: The size of the original image to be processed cannot exceed 32 MB, its width and height cannot exceed 50,000 pixels, and its total number of pixels cannot exceed 250 million. The width and height of the processed image cannot exceed 50,000 pixels. For animated images, the product of the original image's width, height, and number of frames cannot exceed 250 million. Maximum number of frames (for GIF): 300 |
| Supports JPG, BMP, GIF, PNG, WebP (default value: format of the input image). Converting GIF into HEIF is not supported. | |
| Supports JPG and WebP only. | |
| Supports JPG only. If the format of the output image is not JPG, this parameter will be ignored. | |
| Performs multiple operations on an image in sequence. Currently, up to 10 layers of pipelines are supported. | |
Advanced features | Guetzli Image Compression | Supports only JPG images whose quality is larger than 70 and the total number of pixels is smaller than 4 million. |
| Image Advanced Compression | Format: supports converting JPG, PNG, GIF, WebP, and other formats into TPG or HEIF. Size limit: The size of the original image to be processed cannot exceed 32 MB, its width and height cannot exceed 50,000 pixels, and its total number of pixels cannot exceed 250 million. The width and height of the processed image cannot exceed 50,000 pixels. For animated images, the product of the original image's width, height, and number of frames cannot exceed 250 million. Maximum number of frames (for GIF): 300 |
| Blind watermarking | Adding blind watermark to animated images such as GIF is not supported. Width or height of the image watermark cannot be larger than 1/8 of the input image. A text watermark can contain only numbers (0−9) and English letters (A−Z, a−z). |
Media processing | Video Processing | Supports AVI, MP4, MKV, FLV, HLS, TS, MP3, AAC, FLAC, and more. |
Content moderation | Image Moderation | Supports PNG, JPEG, JPG, BMP, WebP, and GIF. An image cannot be larger than 10 MB. To ensure recognition accuracy, the image resolution should not be lower than 256x256. |
| Video Moderation | Supports MP4, AVI, MKV, WMV, RMVB, FLV, and M3U8. A video cannot be larger than 5 GB and the number of frames cannot be more than 100 thousand. |
| Audio Moderation | Supports MP3, WAV, AAC, FLAC, AMR, 3GP, M4A, WMA, OGG, and APE. Audio bitrate: 128−256 kbps Audio size: smaller than 600 MB Maximum duration: 3 hours |
| Text Moderation | Supports TXT files not larger than 1 MB only. |
Content recognition | QR Code Recognition、Image Label Recognition | Supports PNG, JPEG, and JPG. An image cannot be larger than 3 MB and the resolution should be higher than 50x50. |
| Speech Recognition | Supports Mandarin, English, and Cantonese. Supported format: WAV, MP3, M4A An audio file cannot be longer than 5 hours or larger than 512 MB. |
| Face Filter | A Base64-encoded image cannot be larger than 5 MB. PNG, JPG, JPEG, and BMP are supported, while GIF is not supported. To use portrait segmentation, the image resolution must be lower than 2000×2000. |
| ID Card Recognition | A Base64-encoded image cannot be larger than 7 MB, and a resolution higher than 500×800 is recommended. Supported formats are PNG, JPG, JPEG, and BMP. It is recommended that the ID card should occupy over 2/3 of the image. |
File preview | File Preview | The currently supported input file formats include the following: Demonstration files: pptx, ppt, pot, potx, pps, ppsx, dps, dpt, pptm, potm, ppsm. Text files: doc, dot, wps, wpt, docx, dotx, docm, dotm. Spreadsheet files: xls, xlt, et, ett, xlsx, xltx, csv, xlsb, xlsm, xltm, ets. Other format files: pdf, lrc, c, cpp, h, asm, s, java, asp, bat, bas, prg, cmd, rtf, txt, log, xml, htm, html. Documents up to 200 MB and 5,000 pages can be previewed. |
Intelligent Search-MetaInsight | Dataset | The maximum number of datasets per APPID is 100. When an asynchronous export query task is created, the default concurrency per dataset is 5, and the maximum number of queued tasks is 10. The maximum data volume of a dataset using the Basic Information Template is 100 million files. When a dataset is bound to more than 100 million files from data sources, new index creation will be stopped. For a dataset using the Image Search or Face Search Template with the Basic Specification, the maximum data volume per dataset is 500,000. For a dataset using the Image Search or Face Search Template with the Standard Specification, the maximum data volume per dataset is 100 million. The maximum number of data sources that can be bound to a dataset is 100. A single bucket can have up to 100 binding relationships. The concurrency for binding datasets to data sources through the API is 1 QPS. Encrypted files in data sources cannot be read. |
| Metadata Index | The concurrency for creating indexes through API is 20 QPS. |
| Scalar search | The default concurrency for simple queries is 20 QPS. By default, a single query returns information about 100 files, with a maximum of 200 files. For a single query using aggregation parameters, 100 aggregation results are returned by default, with a maximum of 2,000 aggregation results. A maximum of 100 subquery conditions can be set for a single query. The maximum supported nesting depth for a single subquery is 5 levels. |
| Hybrid Search - Image Search | Input image format restrictions: PNG, JPG, and JPEG are supported. GIF and BMP images are not supported. Input image size limits: The Base64-encoded image must not exceed 10 MB, and the image resolution must not exceed 4096px * 4096px. The default concurrency for image search requests is 20 QPS. A single search request can return a maximum of 100 relevant images. |
| Face Search | Input image format restrictions: PNG, JPG, and JPEG are supported. GIF and BMP images are not supported. Input image size limits: The Base64-encoded image must not exceed 10 MB, and the image resolution must not exceed 4096px * 4096px. The default concurrency for face search requests is 20 QPS. A single search can recognize up to 10 faces in an image. Up to 100 related faces can be returned for each face, meaning a maximum of 100 related face images can be returned for a single face image. |
| Metadata filtering | CustomId field limit: 256 bytes at most. CustomLabels field limits: Up to 10 entries. Each Key can be up to 256 bytes, and each value can be up to 4 KB. COSUserMeta field limits: Up to 10 entries. Each entry is limited to 2 KB, and the total size cannot exceed 4 KB. COSTagging field limits: Up to 10 entries. Each Key can be up to 128 bytes, and each value can be up to 256 bytes. |
フィードバック