![\"A](\"https://replicate.delivery/xezq/eZyhkR2Z9o1UHSvvcdH87NxJdfdXLrl0KBC83JHoyDEpqlsUA/out-0.webp\")
\n\nWith Buzz, you can generate full-body character shots with symmetrical compositions and cool retro color palettes. The model excels at creating balanced, stylized imagery that maintains a cohesive cartoon aesthetic.\n\nBuzz offers flexibility with two generation modes: a dev model that performs best with around 28 inference steps for higher quality, and a schnell model that needs only 4 steps for faster generation. You can also use it for image-to-image transformations or inpainting with masks.\n\n## Use cases\n\n- Create consistent cartoon-style illustrations\n- Design retro-inspired character concepts\n- Generate symmetrical scene compositions\n- Produce stylized character portraits\n- Create promotional artwork with consistent styling\n- Generate visual assets for animation projects\n- Design character sheets with full-body views\n- Create social media graphics with retro aesthetics\n\n## Try it yourself\n\nGet started with Buzz using the JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"replicate/buzz\",\n {\n input: {\n prompt: \"a retro cartoon style woman with cool retro colors standing on a skateboard in a symmetrical scene\",\n model: \"dev\",\n num_inference_steps: 28,\n guidance_scale: 3,\n aspect_ratio: \"1:1\",\n num_outputs: 1\n }\n }\n);\n\nconsole.log(output);\n```\n\nCheck out [Buzz on Replicate](https://replicate.com/replicate/buzz) to start generating your own retro cartoon artwork today.","taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-15T20:44:24.000Z","updatedAt":"2025-05-15T20:44:24.000Z"},{"model":"wavespeedai/wan-2.1-i2v-480p","classification":{"summary":"Turns still images into smooth videos with text prompts","inputTypes":["image","text"],"outputTypes":["video"],"task":"image-to-video","useCases":["Add motion to product photos","Create talking head videos from portraits","Turn landscape photos into nature videos","Animate characters from concept art","Create visual effects from still images","Turn profile pictures into dynamic content","Generate video ads from product images","Create animated social media content","Turn paintings into short animations","Create educational visual demonstrations"],"blogPost":"---\ntitle: Animate any image with Wan 2.1 Image-to-Video\nauthors:\n - blog-o-matic\nintro: Transform your still images into smooth, realistic videos with the powerful Wan 2.1 Image-to-Video model.\npublishedAt: 2023-09-25\n---\n\nThe Wan 2.1 Image-to-Video model represents a breakthrough in transforming static images into engaging, fluid videos. By combining the power of a 14B parameter model with acceleration optimizations, this model makes high-quality video generation accessible and practical.\n\nYou can try the model yourself at [replicate.com/wan-2-1-i2v-480p](https://replicate.com/wan-2-1-i2v-480p).\n\n## See it in action\n\nHere's a woman transformed into a talking video:\n\n\n\nAnd here's a cat animated to knead on a laptop:\n\n\n\nFor more creative effects, you can even use LoRA weights to get specialized animations - like this squishable cat effect:\n\n\n\n## What can you do with it?\n\nThe Wan 2.1 Image-to-Video model opens up a world of creative possibilities:\n\n- Turn product photos into dynamic showcases\n- Create animated portraits from headshots\n- Generate nature videos from landscape photos\n- Bring illustrations and concept art to life\n- Create attention-grabbing social media content\n- Add motion to educational diagrams\n- Transform paintings into short animations\n- Generate dynamic visual effects\n\n## How to use it\n\nIt's easy to integrate Wan 2.1 into your applications with just a few lines of code:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"wan-2-1-i2v-480p\",\n {\n input: {\n image: \"https://example.com/your-image.jpg\",\n prompt: \"A cat is sitting on a laptop, it is kneading\",\n max_area: \"832x480\",\n num_frames: 81,\n sample_shift: 3,\n sample_steps: 30,\n frames_per_second: 16,\n sample_guide_scale: 5\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model gives you fine-grained control over parameters like frame count, sample steps, and guide scale so you can balance quality and generation speed according to your needs. For specialized animations, you can even load LoRA weights to get effects like the \"squish\" animation shown above.\n\nWith Wan 2.1, you're just one API call away from bringing your images to life with fluid, high-quality animation."},"createdAt":"2025-05-15T21:23:01.000Z","updatedAt":"2025-05-15T21:23:01.000Z"},{"model":"pixverse/pixverse-v4.5","classification":{"summary":"Creates high-quality videos from text prompts or images","inputTypes":["text","image"],"outputTypes":["video"],"task":"text-to-video","categories":["video","generative","animation","motion","prompt-conditioning"],"useCases":["Create short video clips from descriptions","Transform still images into moving videos","Generate product showcases for marketing","Create social media content quickly","Visualize stories and concepts","Make animations from text prompts","Create educational visual content","Prototype video ideas before production","Design motion graphics from descriptions","Generate scene transitions for presentations"],"blogPost":"---\ntitle: Pixverse v4.5 - Create smooth, high-quality videos from text\nauthors:\n - blog-o-matic\nintro: Generate 5-8 second videos with enhanced motion and complex actions\npublishedAt: 2023-12-01\n---\n\nPixverse v4.5 takes text-to-video generation to the next level, allowing you to quickly create 5-8 second videos with enhanced motion fluidity and improved prompt coherence.\n\nThis latest model excels at handling complex actions and can generate videos at various resolutions up to 1080p. You can start from a text prompt alone or provide an image for the first frame to guide the video generation.\n\n\n\n## Key features\n\n- Generate videos at 540p, 720p, or 1080p resolution\n- Choose between 5 or 8 second durations\n- Normal or smooth motion modes (smooth available for 5 second videos)\n- Multiple aspect ratio options (16:9, 9:16, etc.)\n- First frame image input support\n- Last frame image targeting for transitions\n\n\n\n## Use cases\n\n- Create short-form content for social media\n- Visualize creative concepts quickly\n- Generate product demonstrations\n- Produce educational animations\n- Create engaging marketing materials\n- Prototype video ideas before full production\n\n\n\n## Try it yourself\n\nHere's how to generate videos with Pixverse v4.5 using the Replicate JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"pixverse/pixverse-v4.5\",\n {\n input: {\n prompt: \"a snow leopard is walking carefully, snowy landscape scene at twilight\",\n quality: \"720p\",\n duration: 5,\n motion_mode: \"normal\",\n aspect_ratio: \"16:9\",\n negative_prompt: \"\"\n }\n }\n);\n\nconsole.log(output);\n```\n\nExperiment with different prompts, resolutions, and motion settings to create the perfect video for your needs. [Try Pixverse v4.5 on Replicate](https://replicate.com/pixverse/pixverse-v4.5) today!","taskSummary":"Text-to-video models can be used in any application that requires generating consistent sequence of images from text. "},"createdAt":"2025-05-17T18:38:24.000Z","updatedAt":"2025-05-17T18:38:24.000Z"},{"model":"openai/dall-e-3","classification":{"summary":"AI system that creates realistic images from text descriptions","inputTypes":["text"],"outputTypes":["image"],"task":"text-to-image","categories":["image","generative","art","creative","prompt-conditioning"],"useCases":["Create custom artwork from text descriptions","Generate product visuals for marketing","Visualize architectural concepts from descriptions","Design characters for games or stories","Produce educational illustrations for learning materials","Create visual aids for presentations","Generate social media content","Visualize interior design concepts","Create custom book cover art","Prototype visual concepts quickly"],"blogPost":"---\ntitle: Create stunning images with DALL-E 3\nauthors:\n - blog-o-matic\nintro: DALL-E 3 generates highly realistic images and art from natural language descriptions\npublishedAt: 2023-10-03\n---\n\nOpenAI's DALL-E 3 model takes text-to-image generation to new heights, creating detailed, accurate, and visually stunning images from natural language descriptions. This powerful model understands complex prompts and delivers high-quality visuals that closely match your creative vision.\n\nYou can try DALL-E 3 now on [Replicate](https://replicate.com/openai/dall-e-3).\n\nDALL-E 3 represents a significant advancement in AI image generation. The model is trained to understand nuanced textual descriptions and translate them into visually coherent images that faithfully represent what you're describing. Whether you need photorealistic images, stylized illustrations, abstract concepts, or detailed scenes, DALL-E 3 can bring your ideas to life.\n\n## Example use cases\n\n- Create custom illustrations for blog posts and articles\n- Generate concept art for films, games, or products\n- Design marketing materials with specific themes or styles\n- Visualize architectural or interior design concepts\n- Create educational content with visual explanations\n- Generate unique social media visuals\n- Prototype visual ideas before manual creation\n- Produce art in various styles for creative projects\n- Create custom avatars or character designs\n- Visualize scientific or abstract concepts\n\n## How to use DALL-E 3\n\nYou can run DALL-E 3 using the Replicate JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"openai/dall-e-3\",\n {\n input: {\n prompt: \"A majestic lion jumping from a big rock at sunset\"\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model accepts a text prompt describing the image you want to create. You can be as specific as you like with details about subject, setting, lighting, style, color scheme, and more. The more detailed your prompt, the more likely the output will match your vision.","taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T19:24:32.000Z","updatedAt":"2025-05-17T19:24:32.000Z"},{"model":"nvidia/sana-sprint-1.6b","classification":{"summary":"Fast text-to-image generation model with one-step diffusion","inputTypes":["text","parameters"],"outputTypes":["image"],"task":"text-to-image","categories":["diffusion","image","text-to-image","prompt-conditioning","ai-art"],"useCases":["Generate creative concept art","Create illustrations from text descriptions","Design character portraits","Visualize interior design concepts","Produce fantasy landscapes","Create product mockups","Generate storyboard frames","Visualize architectural concepts","Create social media visuals","Design book cover concepts"],"blogPost":"---\ntitle: SANA-Sprint: Lightning-fast image generation in just 2 steps\nauthors:\n - blog-o-matic\nintro: Generate high-quality images from text prompts incredibly fast with SANA-Sprint's one-step diffusion model\npublishedAt: 2023-11-01\n---\n\nMeet SANA-Sprint, a remarkable breakthrough in diffusion model technology that generates impressive images in just 1-2 inference steps. Unlike traditional diffusion models that require dozens of steps, SANA-Sprint uses Continuous-Time Consistency Distillation to produce high-quality images almost instantly.\n\n[SANA-Sprint](https://replicate.com/nvidia/sana-sprint-1.6b) is a lightweight 1.6B parameter model developed by NVIDIA that prioritizes speed without compromising quality. This efficiency makes it perfect for applications requiring quick visual generation.\n\n## See SANA-Sprint in action\n\nGenerate fantastical scenes from simple text prompts:\n\n![\"a](\"https://replicate.delivery/xezq/tbAO9mMLiSKFMVNefs91HKfNwsW6mPLsDm4whM4FoPeMhK6RB/output.jpg\")
\n\nCreate charming character portraits with just a few words:\n\n![\"Peach-faced](\"https://replicate.delivery/xezq/LwfLdLkwAr08KS3sTzBA9eHemndetHWvhQqqyReJj2A2k3xjC/output.png\")
\n\nGenerate stunning landscapes with photographic detail:\n\n![\"Close-up](\"https://replicate.delivery/xezq/YWbRNBoTeflNaU8KpACo1EJNGm8MQhQvCReMUF6mJDbl0d8oA/output.png\")
\n\nOr visualize imaginative interior designs:\n\n![\"a](\"https://replicate.delivery/xezq/diSPTDnOzCK6BBpeaseS2suzffG1aVXKPPWADW0VYYr9z54RB/output.png\")
\n\n## How to use SANA-Sprint\n\nSANA-Sprint is easy to integrate into your projects. Here's a simple example using JavaScript:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"nvidia/sana-sprint-1.6b\",\n {\n input: {\n prompt: \"a photograph of an astronaut riding a horse\",\n width: 1024,\n height: 1024,\n guidance_scale: 4.5,\n inference_steps: 2\n }\n }\n);\n\nconsole.log(output);\n```\n\n## Use cases\n\nSANA-Sprint excels at a variety of creative and practical applications:\n\n- Creating concept art for games and films\n- Generating illustrations for books or articles\n- Designing unique characters for animation\n- Visualizing interior design concepts\n- Producing fantasy landscapes and environments\n- Creating product mockups\n- Generating storyboard frames\n- Visualizing architectural concepts\n- Creating social media visuals\n- Designing book cover concepts\n\nWith its remarkable speed and quality, SANA-Sprint represents a significant advancement in text-to-image generation technology. Try it today and experience the power of one-step diffusion for yourself.","taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T19:26:14.000Z","updatedAt":"2025-05-17T19:26:14.000Z"},{"model":"easel/advanced-face-swap","classification":{"summary":"Face swap tool that transfers faces onto target images","inputTypes":["image"],"outputTypes":["image"],"task":"image-to-image","categories":["face-swap","image-editing","identity-transfer","style-transfer","portrait-manipulation"],"useCases":["Create fun celebrity face swaps","Try on different hairstyles virtually","Place yourself in historical photos","Transform portrait photos with new identities","Create personalized movie poster images","Put your face in famous paintings","Generate avatars with your likeness","Swap faces for creative photography projects","Create custom memes with face swaps","Replace actors in film screenshots"],"blogPost":"---\ntitle: Advanced Face Swap - Seamlessly Transfer Faces to Target Images\nauthors:\n - blog-o-matic\nintro: Transfer faces between images with natural results and hair style control\npublishedAt: 2023-09-21\n---\n\nThe [Advanced Face Swap model](https://replicate.com/easel/advanced-face-swap) brings professional-grade face swapping capabilities to anyone working with portrait images. This powerful tool lets you transplant one or two faces onto a target image with natural-looking results.\n\n![\"Advanced](\"https://replicate.delivery/xezq/fsV6HOzRAWTuVKvdglKmn0q3TNwfVqeThvYp3gREUT1boX9oA/tmpxrxhromb.jpg\")
\n\nWhat makes this model particularly useful is its ability to control whose hairstyle is preserved in the result. You can choose to keep the hairstyle from either the source face or the target image, giving you more creative control over the final output.\n\nThe model also supports swapping two faces simultaneously, perfect for creating couple portraits or group photos. You can optionally specify gender descriptions to help the model create more accurate swaps for multiple users.\n\n## Use Cases\n\nAdvanced Face Swap opens up numerous creative possibilities:\n\n- Create personalized movie posters with your face\n- Try on different hairstyles virtually\n- Insert yourself into historical photos\n- Generate custom avatars for gaming or social media\n- Create fun celebrity face swaps\n- Design personalized greeting cards\n- Make custom marketing materials with consistent brand faces\n- Create engaging social media content\n\n## How to Use It\n\nHere's how to use Advanced Face Swap with the JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"easel/advanced-face-swap\",\n {\n input: {\n swap_image: \"https://example.com/your-face.jpg\",\n target_image: \"https://example.com/target-image.jpg\",\n hair_source: \"target\" // or \"user\" to keep hair from source image\n }\n }\n);\n\nconsole.log(output);\n```\n\nFor a two-person swap, simply add the second face image and optional gender descriptions:\n\n```javascript\nconst output = await replicate.run(\n \"easel/advanced-face-swap\",\n {\n input: {\n swap_image: \"https://example.com/face1.jpg\",\n swap_image_b: \"https://example.com/face2.jpg\",\n target_image: \"https://example.com/target-image.jpg\",\n user_gender: \"a woman\",\n user_b_gender: \"a man\"\n }\n }\n);\n```\n\nThe model will return a URL to the generated image with the swapped face(s).","taskSummary":"Image-to-image is the task of transforming an input image through a variety of possible manipulations and enhancements, such as super-resolution, image inpainting, colorization, and more."},"createdAt":"2025-05-17T19:26:38.000Z","updatedAt":"2025-05-17T19:26:38.000Z"},{"model":"meta/llama-4-scout-instruct","classification":{"summary":"A 17B parameter language model that generates text","inputTypes":["text"],"outputTypes":["text"],"task":"text-generation","categories":["text","language-model","chat","ai-assistant","instruction-following"],"useCases":["Draft emails and messages","Answer questions on various topics","Summarize long documents","Generate creative writing","Help with problem solving","Create structured content","Explain complex concepts","Assist with coding tasks","Brainstorm ideas","Draft professional documents"],"blogPost":"---\ntitle: Meta's Llama-4-Scout-Instruct brings powerful text generation to Replicate\nauthors:\n - blog-o-matic\nintro: A 17 billion parameter language model with mixture-of-experts architecture for efficient instruction following\npublishedAt: 2023-06-15\n---\n\nMeta has released Llama-4-Scout-Instruct, a powerful new language model available on Replicate. At 17 billion parameters and built with a mixture-of-experts architecture featuring 16 experts, this model delivers impressive text generation capabilities while maintaining computational efficiency.\n\nCheck out the model at [https://replicate.com/meta/llama-4-scout-instruct](https://replicate.com/meta/llama-4-scout-instruct)\n\nLlama-4-Scout-Instruct is designed to follow instructions accurately and generate helpful, coherent responses to prompts. The mixture-of-experts approach means the model can activate only the most relevant neural pathways for each query, allowing it to match the capabilities of much larger models while using fewer computational resources.\n\n## What can you do with Llama-4-Scout-Instruct?\n\n- Generate conversational AI responses\n- Draft content like emails, stories, and articles\n- Answer factual questions\n- Explain complex concepts\n- Summarize long documents\n- Assist with creative writing\n- Help with problem-solving\n- Generate structured content like lists and tables\n- Provide step-by-step instructions\n- Translate between ideas and formal writing\n\n## Running the model\n\nHere's how to use Llama-4-Scout-Instruct with the Replicate JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"meta/llama-4-scout-instruct\",\n {\n input: {\n prompt: \"Write a short story about a robot who discovers poetry.\",\n max_tokens: 512,\n temperature: 0.7,\n top_p: 0.9,\n presence_penalty: 0,\n frequency_penalty: 0\n }\n }\n);\n\nconsole.log(output.join(\"\"));\n```\n\nBy adjusting parameters like temperature and top_p, you can control the creativity and randomness of the model's outputs. Lower values produce more deterministic and focused responses, while higher values enable more creative and diverse outputs.\n\nLlama-4-Scout-Instruct represents an exciting advancement in language model technology, bringing powerful capabilities to developers and creators on Replicate.","taskSummary":"Generating text is the task of generating new text given another text. These models can, for example, fill in incomplete text or paraphrase."},"createdAt":"2025-05-17T19:26:58.000Z","updatedAt":"2025-05-17T19:26:58.000Z"},{"model":"meta/llama-4-maverick-instruct","classification":{"summary":"Powerful text generation model with 17B parameters.","inputTypes":["text"],"outputTypes":["text"],"task":"text-generation","categories":["language","text","generative","instruct","conversational"],"useCases":["Conversational AI for customer support","Creative writing assistance","Content generation for marketing","Technical document drafting","Educational tutoring and explanations","Summarizing long texts","Answering questions with detailed responses","Brainstorming ideas for projects","Translating complex concepts into simple language","Writing code snippets and explanations"],"blogPost":"---\ntitle: Llama 4 Maverick Instruct - Meta's 17B parameter language model\nauthors:\n - blog-o-matic\nintro: Llama 4 Maverick Instruct brings powerful text generation capabilities with a mixture-of-experts architecture.\npublishedAt: 2024-06-01\n---\n\nMeta has released Llama-4-Maverick-Instruct, an impressive 17 billion parameter language model that employs a mixture-of-experts architecture with 128 experts. This design allows the model to achieve exceptional performance while maintaining reasonable computational requirements.\n\nThe model is available on [Replicate](https://replicate.com/meta/llama-4-maverick-instruct) and ready for integration into your applications.\n\n## What makes Llama 4 Maverick special?\n\nThe Maverick variant of Llama 4 uses a sophisticated mixture-of-experts (MoE) architecture, which means it contains 128 specialized neural networks that activate for different types of queries. Rather than using all parameters for every request, the model selectively engages only the most relevant experts for each task, making it both more efficient and more capable.\n\n## Use cases\n\nLlama 4 Maverick Instruct excels at a variety of text generation tasks including:\n\n- Conversational AI and virtual assistants\n- Content creation and creative writing\n- Answering complex questions with nuanced responses\n- Summarizing long documents\n- Explaining complex topics in simple terms\n- Drafting emails and professional communications\n- Problem-solving across different domains\n- Story and narrative generation\n\n## Try it yourself\n\nIntegrating Llama 4 Maverick Instruct into your application is straightforward with Replicate's JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"meta/llama-4-maverick-instruct\",\n {\n input: {\n prompt: \"Explain how mixture-of-experts architecture works in language models.\",\n max_tokens: 1024,\n temperature: 0.6,\n top_p: 1,\n presence_penalty: 0,\n frequency_penalty: 0\n }\n }\n);\n\nconsole.log(output.join(\"\"));\n```\n\n## Model details\n\nCreated by Meta, Llama 4 Maverick Instruct represents a significant advancement in the company's open language model efforts. The mixture-of-experts approach allows this 17B parameter model to achieve performance comparable to much larger dense models while being more computationally efficient.\n\nThis instruct-tuned version is specifically optimized for following user instructions and generating helpful, accurate responses across a wide range of topics and tasks.","taskSummary":"Generating text is the task of generating new text given another text. These models can, for example, fill in incomplete text or paraphrase."},"createdAt":"2025-05-17T19:27:22.000Z","updatedAt":"2025-05-17T19:27:22.000Z"},{"model":"ibm-granite/granite-3.3-8b-instruct","classification":{"summary":"An 8B parameter language model for text generation","inputTypes":["text"],"outputTypes":["text"],"task":"text-generation","categories":["text","language-model","reasoning","instruction-following","long-context"],"useCases":["Answer complex questions with detailed reasoning","Generate creative writing and stories","Summarize long documents or articles","Assist with drafting emails or messages","Explain complex topics simply","Generate code snippets and explanations","Write essays and academic content","Create conversational AI responses","Draft business documents and proposals","Translate concepts across knowledge domains"],"blogPost":"---\ntitle: Granite 3.3 8B Instruct - Powerful reasoning in a compact model\nauthors:\n - blog-o-matic\nintro: A compact yet powerful language model optimized for reasoning and instruction following with 128K context length\npublishedAt: 2023-11-20\n---\n\nGranite-3.3-8B-Instruct delivers impressive reasoning capabilities in a relatively compact 8 billion parameter package. This model from IBM offers extended context length of 128K tokens, making it suitable for processing lengthy documents while maintaining coherent responses.\n\nThe model excels at instruction following and complex reasoning tasks, making it a versatile option for applications requiring thoughtful text generation within a more efficient compute footprint than larger models.\n\n## Capabilities and use cases\n\nWith Granite-3.3-8B-Instruct, you can:\n\n- Answer complex questions with detailed reasoning\n- Generate creative content like stories and poems\n- Summarize long documents while preserving key information\n- Provide explanations of complex topics in simple terms\n- Draft professional communications and documents\n- Generate code with explanations\n- Create conversational AI responses\n\nThe model's 128K context window allows it to process much more information than standard models, enabling it to maintain coherence across lengthy inputs.\n\n## Using Granite-3.3-8B-Instruct\n\nYou can integrate this model into your applications using the Replicate API. Here's a simple example using JavaScript:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"ibm-granite/granite-3.3-8b-instruct\",\n {\n input: {\n prompt: \"Explain the concept of perplexity in language models and why it matters.\",\n temperature: 0.6,\n max_tokens: 500,\n top_p: 0.9,\n top_k: 50\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model provides several parameters to fine-tune its output:\n\n- **temperature**: Controls randomness (lower values are more deterministic)\n- **top_p** and **top_k**: Define token sampling strategies\n- **max_tokens**: Limits the length of generated responses\n- **system_prompt**: Sets the overall behavior and context\n\nYou can learn more and try the model at [https://replicate.com/ibm-granite/granite-3.3-8b-instruct](https://replicate.com/ibm-granite/granite-3.3-8b-instruct).","taskSummary":"Generating text is the task of generating new text given another text. These models can, for example, fill in incomplete text or paraphrase."},"createdAt":"2025-05-17T19:27:41.000Z","updatedAt":"2025-05-17T19:27:41.000Z"},{"model":"kwaivgi/kling-v2.0","classification":{"summary":"Text-to-video generation model creating short HD videos","inputTypes":["text","image","number"],"outputTypes":["video"],"task":"text-to-video","categories":["video","animation","text-to-video","creative","cinematic"],"useCases":["Create short animated stories from text","Generate product demo videos","Visualize screenplay ideas quickly","Create dynamic social media content","Turn concept descriptions into visual demonstrations","Prototype video advertisements from scripts","Generate educational video animations","Create visual transitions between scenes","Turn text prompts into cinematic clips","Visualize creative concepts without filming"],"blogPost":"---\ntitle: Generate short HD videos from text with Kling v2.0\nauthors:\n - blog-o-matic\nintro: Create impressive 5-10 second videos in 720p from text descriptions\npublishedAt: 2023-05-15\n---\n\nText-to-video technology continues to evolve rapidly, and Kling v2.0 represents an impressive step forward in this space. This model generates high-quality 5-10 second videos in 720p resolution based on your text prompts.\n\nThe model was created by kwaivgi and is available at [https://replicate.com/kwaivgi/kling-v2.0](https://replicate.com/kwaivgi/kling-v2.0).\n\n\n\n\n\n\n\n## What you can do with Kling v2.0\n\nKling v2.0 opens up a range of creative possibilities for video generation:\n\n- Create short animated stories from text descriptions\n- Generate product demos and visualizations\n- Produce social media content quickly\n- Prototype video concepts before production\n- Visualize creative ideas without filming\n- Generate educational animations\n- Create dynamic transitions between scenes\n\n## How it works\n\nYou can control several parameters to customize your video output:\n\n- **Prompt**: Text description of what you want to see in the video\n- **Duration**: Length of the video (5 or 10 seconds)\n- **CFG Scale**: Controls flexibility in generation (lower values allow more creativity)\n- **Start Image**: Optional first frame to begin your video from\n- **Aspect Ratio**: Set the dimensions of your video\n- **Negative Prompt**: Specify elements you don't want in the video\n\n## Try it yourself\n\nHere's how you can use Kling v2.0 with the JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"kwaivgi/kling-v2.0\",\n {\n input: {\n prompt: \"Aerial view of the city with flying vehicles navigating through neon-lit skyscrapers\",\n duration: 5,\n cfg_scale: 0.5,\n aspect_ratio: \"16:9\",\n negative_prompt: \"\"\n }\n }\n);\n\nconsole.log(output);\n```\n\nKling v2.0 represents an exciting development in the text-to-video space, making it easier than ever to quickly generate high-quality short videos from simple text prompts.","taskSummary":"Text-to-video models can be used in any application that requires generating consistent sequence of images from text. "},"createdAt":"2025-05-17T19:28:06.000Z","updatedAt":"2025-05-17T19:28:06.000Z"},{"model":"topazlabs/video-upscale","classification":{"summary":"Upscales videos to higher resolution and framerate","inputTypes":["video"],"outputTypes":["video"],"task":"image-to-image","categories":["video","upscaling","super-resolution","enhancement","video-processing"],"useCases":["Improve low resolution video footage","Enhance old or degraded videos","Upgrade home videos to 4K","Convert low-fps content to smoother video","Prepare videos for professional presentation","Upscale video game recordings","Enhance security camera footage","Improve streaming video quality","Restore archival film footage","Upgrade social media video content"],"blogPost":"---\ntitle: Enhance your videos with Topaz Labs' Video Upscale model\nauthors:\n - blog-o-matic\nintro: Transform low-resolution videos into high-quality 4K content with higher framerates\npublishedAt: 2023-07-21\n---\n\nTopaz Labs brings their professional-grade video enhancement technology to Replicate with their Video Upscale model. This powerful tool lets you transform low-resolution videos into crisp, high-definition content while also increasing the framerate for smoother playback.\n\nThe model can upscale your videos to 4K resolution and increase framerates up to 60fps, breathing new life into older footage or enhancing new content for professional applications.\n\n\n\n## What you can do with Video Upscale\n\nThis model is perfect for a variety of video enhancement needs:\n\n- Convert standard definition to 4K resolution\n- Increase frame rates for smoother playback\n- Restore and enhance archival footage\n- Improve quality of streaming content\n- Enhance security camera recordings\n- Upscale video game footage\n- Prepare content for professional presentations\n\n## Using the model\n\nIntegrating Video Upscale into your workflow is straightforward with the Replicate API:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"topazlabs/video-upscale\",\n {\n input: {\n video: \"https://example.com/your-video.mp4\",\n target_fps: 60,\n target_resolution: \"4k\"\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model accepts a video file URL as input, along with parameters for your desired framerate and resolution. After processing, you'll receive a URL to your enhanced video file.\n\nCheck out [Topaz Labs' Video Upscale model](https://replicate.com/topazlabs/video-upscale) on Replicate to start enhancing your videos today.","taskSummary":"Image-to-image is the task of transforming an input image through a variety of possible manipulations and enhancements, such as super-resolution, image inpainting, colorization, and more."},"createdAt":"2025-05-17T19:28:23.000Z","updatedAt":"2025-05-17T19:28:23.000Z"},{"model":"topazlabs/image-upscale","classification":{"summary":"Professionally enhances and upscales images with various models","inputTypes":["image"],"outputTypes":["image"],"task":"image-to-image","categories":["upscaling","super-resolution","image","restoration","enhancement"],"useCases":["Enhance low-resolution photos","Upscale digital art for printing","Improve image quality for social media","Restore old family photographs","Sharpen text in scanned documents","Create high-resolution prints from small images","Fix blurry photos from smartphones","Enhance details in landscape photography","Improve quality of product photography","Rescue low-quality images for professional use"],"blogPost":"---\ntitle: Professional image upscaling with Topaz Labs\nauthors:\n - blog-o-matic\nintro: Transform low-resolution images into crisp, high-quality visuals with Topaz Labs' professional upscaling technology\npublishedAt: 2023-10-25\n---\n\nNeed to transform your blurry, low-resolution images into sharp, detailed visuals? The image-upscale model from Topaz Labs brings professional-grade image enhancement technology to Replicate.\n\nThis model uses advanced AI to intelligently upscale images by up to 4x their original size while preserving and enhancing details that typical resizing methods would lose. With specialized enhancement models for different image types, you can get optimal results for various content.\n\n## See it in action\n\n### Digital art enhancement\n\n\n\n### Low-resolution image enhancement with face detection\n\n\n\n## Features\n\n- Multiple enhancement models tailored for specific types of content\n- Up to 4x upscaling factor\n- Special face enhancement technology\n- Subject detection for foreground emphasis\n- Adjustable enhancement parameters for fine-tuned results\n\n## Use cases\n\n- Convert low-resolution images to high-quality prints\n- Enhance old photos for digital preservation\n- Improve the quality of web images\n- Upscale digital art for commercial use\n- Sharpen text in document scans\n- Create higher resolution versions of small product photos\n- Rescue low-quality historical images\n- Enhance social media visuals for better engagement\n\n## Try it yourself\n\nHere's how to use this model with the Replicate JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"topazlabs/image-upscale\",\n {\n input: {\n image: \"https://example.com/your-image.jpg\",\n enhance_model: \"CGI\", // Options: Standard V2, Low Resolution V2, CGI, High Fidelity V2, Text Refine\n output_format: \"jpg\",\n upscale_factor: \"4x\", // Options: 2x, 4x\n face_enhancement: true,\n subject_detection: \"Foreground\", // Options: None, Foreground\n face_enhancement_strength: 0.8,\n face_enhancement_creativity: 0.3\n }\n }\n);\n\nconsole.log(output);\n```\n\nExplore and enhance your images with this professional upscaling tool from Topaz Labs at [replicate.com/topazlabs/image-upscale](https://replicate.com/topazlabs/image-upscale).","taskSummary":"Image-to-image is the task of transforming an input image through a variety of possible manipulations and enhancements, such as super-resolution, image inpainting, colorization, and more."},"createdAt":"2025-05-17T19:28:47.000Z","updatedAt":"2025-05-17T19:28:47.000Z"},{"model":"black-forest-labs/flux-1.1-pro-ultra-finetuned","classification":{"summary":"Fine-tunable text-to-image model supporting custom fine-tunes","inputTypes":["text","image","boolean","integer","number","string"],"outputTypes":["image"],"task":"text-to-image","categories":["diffusion","image","prompt-conditioning","style-transfer","fine-tuning"],"useCases":["Create custom branded product images","Generate art in specific styles","Design concept illustrations with personalized aesthetics","Produce consistent character artwork","Create custom marketing visuals","Design unique digital assets","Generate product mockups","Create personalized illustrations","Design custom thumbnails","Produce stylized social media content"],"blogPost":"---\ntitle: Flux 1.1 Pro Ultra - Fine-tunable text-to-image model with 4MP resolution\nauthors:\n - blog-o-matic\nintro: Generate high-quality images with custom fine-tuned styles using FLUX 1.1 Pro Ultra\npublishedAt: 2023-12-15\n---\n\nFlux 1.1 Pro Ultra is a powerful text-to-image model that supports custom fine-tuning, allowing you to generate images in specific styles that you've trained the model to understand. The model supports impressive 4MP resolution and offers a raw mode for more natural-looking, realistic outputs.\n\nCheck out the model at [black-forest-labs/flux-1.1-pro-ultra-finetuned](https://replicate.com/black-forest-labs/flux-1.1-pro-ultra-finetuned).\n\n## What makes it special?\n\nUnlike standard text-to-image models, Flux 1.1 Pro Ultra lets you use custom fine-tunes through a `finetune_id` parameter. This means you can train the model on a specific style, character, or concept, then generate unlimited variations in that style.\n\nHere's an example output from the model using a custom fine-tune:\n\n[](https://replicate.com/black-forest-labs/flux-1.1-pro-ultra-finetuned)\n\nThe model also supports composition guidance through an optional image prompt, allowing you to create images that follow the composition of a reference image while applying your custom style.\n\n## Key features\n\n- Support for custom fine-tunes via `finetune_id`\n- High-resolution 4MP image output\n- Raw mode option for more natural-looking images\n- Aspect ratio control\n- Image prompt support for composition guidance\n- Adjustable safety settings\n\n## Use cases\n\nFlux 1.1 Pro Ultra with fine-tuning can be used for:\n\n- Creating consistent branded visuals for marketing\n- Generating art in a specific artist's style\n- Producing character art with consistent appearances\n- Designing concept art for games or films\n- Creating product mockups in custom styles\n- Generating personalized illustrations\n- Producing design assets with consistent aesthetics\n\n## How to use it\n\nHere's how to run the model using the JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"black-forest-labs/flux-1.1-pro-ultra-finetuned\",\n {\n input: {\n prompt: \"A formula one car in vibrant colors\",\n finetune_id: \"YOUR_FINETUNE_ID\",\n aspect_ratio: \"3:2\",\n raw: false,\n safety_tolerance: 2,\n finetune_strength: 0.75\n }\n }\n);\n\nconsole.log(output);\n```\n\nTo get the most from this model, you'll need to first create a fine-tune using Replicate's Flux Pro fine-tuning model. Once you have your `finetune_id`, you can use it to generate images in that specific style.","taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T19:38:49.000Z","updatedAt":"2025-05-17T19:38:49.000Z"},{"model":"openai/gpt-image-1","classification":{"summary":"Creates high-quality images from text and image inputs","inputTypes":["text","image"],"outputTypes":["image"],"task":"text-to-image","categories":["image","generative","multimodal","style-transfer","prompt-conditioning"],"useCases":["Create product photos from descriptions","Generate artwork from text prompts","Transform sketches into polished images","Design marketing materials with specific aesthetic","Visualize interior design concepts","Create custom illustrations for content","Generate variations of existing images","Design concept art for games","Create visual assets for presentations","Transform photos with specific styles"],"blogPost":"---\ntitle: OpenAI's GPT-Image-1 Arrives on Replicate\nauthors:\n - blog-o-matic\nintro: Generate high-quality images with OpenAI's text-to-image model on Replicate\npublishedAt: 2023-11-15\n---\n\nOpenAI's GPT-Image-1 is now available on Replicate. This powerful multimodal image generation model creates high-quality images from text descriptions and can even use existing images as input reference.\n\n[Try GPT-Image-1 on Replicate](https://replicate.com/openai/gpt-image-1)\n\n![\"Example](\"https://replicate.delivery/xezq/hEcJpm50O9ouFVu506qf6wAjLKc4MH1oFeRZ4PORmTqVqVnUA/tmpfkxavg8q.png\")
\n\nDeveloped by OpenAI, GPT-Image-1 can handle a variety of image generation tasks. The model requires your own verified OpenAI API key to use.\n\n## What can you do with GPT-Image-1?\n\nGPT-Image-1 excels at creating detailed images from text descriptions. You can use it to:\n\n- Generate custom illustrations for articles and presentations\n- Create product mockups from descriptions\n- Design concept art for creative projects\n- Transform sketches or reference images with specific styles\n- Produce marketing materials with consistent brand aesthetics\n- Visualize interior design concepts\n- Create personalized artwork from text prompts\n- Generate visual assets for websites and apps\n\n## Using GPT-Image-1 with Replicate\n\nYou can run GPT-Image-1 with a few lines of code using the Replicate JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"openai/gpt-image-1\",\n {\n input: {\n prompt: \"A vibrant watercolor painting of a coastal village at sunset\",\n quality: \"auto\",\n aspect_ratio: \"16:9\",\n openai_api_key: \"your-openai-api-key\",\n number_of_images: 1\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model supports various parameters including quality settings, aspect ratio control, and the ability to generate multiple images in a single run. You can also use input images as references for your generations.\n\nGive GPT-Image-1 a try today for your next creative project!","taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T19:39:18.000Z","updatedAt":"2025-05-17T19:39:18.000Z"},{"model":"black-forest-labs/flux-pro-trainer","classification":{"summary":"Fine-tune FLUX 1.1 Pro with custom images","inputTypes":["image","text"],"outputTypes":["text"],"task":"text-to-image","categories":["diffusion","fine-tuning","customization","image-generation","training"],"useCases":["Create custom character generators","Train models on specific art styles","Develop brand-specific image generators","Personalize image generation with family photos","Create custom product visualization tools","Train models on specialized industry assets","Build custom architecture visualization tools","Create personalized avatar generators","Develop custom fashion design systems","Build specialized medical imaging generators"],"blogPost":"---\ntitle: Fine-tune FLUX 1.1 Pro with your own images\nauthors:\n - blog-o-matic\nintro: Train custom FLUX models with your own image datasets\npublishedAt: 2023-10-12\n---\n\nFine-tuning powerful image generation models with your own images just got easier. The FLUX Pro Trainer lets you create custom versions of FLUX 1.1 Pro and FLUX 1.1 Pro Ultra by uploading your own image datasets.\n\nThis model from Black Forest Labs enables you to train specialized image generators that reflect your unique style, characters, or brand elements. The process is straightforward: upload your images, specify a trigger word, and the trainer will create a custom fine-tuned model for you to use.\n\n## How it works\n\nThe training process is simple:\n\n1. Gather 10+ images that represent what you want your model to learn\n2. Optionally include captions as .txt files with the same name as your images\n3. Upload the images as a zip file\n4. Choose a unique trigger word that will activate your fine-tuned model\n5. Select your preferred fine-tuning parameters\n6. Get back a unique finetune_id to use with the FLUX inference model\n\nYou can control various aspects of the training process including learning rate, number of training steps, and whether to enable automatic captioning for your images.\n\n## Use cases\n\nWith FLUX Pro Trainer, you can:\n\n- Create custom character generators for game or film development\n- Train models on your own artistic style\n- Develop brand-specific image generators\n- Build specialized industry-specific visual tools\n- Create personalized avatar generators\n- Train models on architectural styles or product designs\n\n## Getting started\n\nHere's how to run the model using the JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate({\n auth: process.env.REPLICATE_API_TOKEN,\n});\n\nconst output = await replicate.run(\n \"black-forest-labs/flux-pro-trainer\",\n {\n input: {\n mode: \"general\",\n priority: \"quality\",\n lora_rank: 32,\n captioning: \"captioning-disabled\",\n input_images: \"https://example.com/my-images.zip\",\n trigger_word: \"MYSTYLE\",\n finetune_type: \"full\",\n training_steps: 300\n }\n }\n);\n\nconsole.log(output); // This will be your finetune_id\n```\n\nOnce you have your finetune_id, you can use it with the FLUX inference model to generate images with your custom model.\n\nVisit [black-forest-labs/flux-pro-trainer](https://replicate.com/black-forest-labs/flux-pro-trainer) to try it yourself and create your own custom FLUX model today.","taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T19:39:39.000Z","updatedAt":"2025-05-17T19:39:39.000Z"},{"model":"zsxkib/step1x-edit","classification":{"summary":"Edits images based on text instructions","inputTypes":["image","text"],"outputTypes":["image"],"task":"image-to-image","categories":["image-editing","prompt-conditioning","diffusion","style-transfer","inpainting"],"useCases":["Replace objects in photos","Change clothing in fashion images","Remove unwanted elements from pictures","Transform image styles","Add objects to existing scenes","Change colors of specific elements","Convert photos into artistic illustrations","Enhance product photos for e-commerce","Create variations of existing images","Modify backgrounds in portraits"],"blogPost":"---\ntitle: Step1X-Edit - Transform your images with text prompts\nauthors:\n - blog-o-matic\nintro: Edit and transform images with simple text instructions using Step1X-Edit\npublishedAt: 2023-11-28\n---\n\nText-guided image editing has never been easier thanks to Step1X-Edit, a powerful model that lets you transform images with natural language instructions.\n\nDeveloped by stepfun-ai, Step1X-Edit allows you to make targeted modifications to images by simply describing what you want to change. Want to replace a leather jacket with a furry pink one? Remove an object? Transform a photo into a Studio Ghibli illustration? Step1X-Edit can handle it all.\n\n## See it in action\n\n\n ![\"Leather](\"https://replicate.delivery/yhqm/DbCq0sv0s2bgGhN2hpyHwQ19tM31ThsekJpqHseeGr1ahJRpA/step1x_edit_output.webp\")
\n\nPrompt: \"Replace the leather jacket with a furry pink jacket, add sunglasses\"
\n\n\n![\"Moth](\"https://replicate.delivery/yhqm/jk5dmx205greBCNJCEwo3Ugexu1TR0GoorXxfrPjUyd14qQpA/step1x_edit_output.webp\")
\n\nPrompt: \"replace moth with a small humming bird\"
\n\n\n![\"Photo](\"https://replicate.delivery/yhqm/VWlZTvnsJbYoENHQonzHvP6ddaLArFzXsdZf7YOdTCvaifnUA/step1x_edit_output.webp\")
\n\nPrompt: \"turn into an illustration in studio ghibli style\"
\n\n## What can you do with Step1X-Edit?\n\nThe model is incredibly versatile and can be used for many creative applications:\n\n- Replace specific objects or elements in photos\n- Add new items to existing scenes\n- Remove unwanted elements from images\n- Change colors and textures\n- Transform photographic styles into artistic illustrations\n- Modify clothing or accessories on people\n- Change backgrounds while preserving foreground subjects\n\n## Try it yourself\n\nYou can start using Step1X-Edit right away with the Replicate JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\nconst output = await replicate.run(\n \"zsxkib/step1x-edit\",\n {\n input: {\n image: \"https://example.com/your-image.jpg\",\n prompt: \"Replace the blue shirt with a red sweater\",\n size_level: 512,\n output_format: \"webp\",\n output_quality: 80\n }\n }\n);\n\nconsole.log(output);\n```\n\nCheck out the [Step1X-Edit model page](https://replicate.com/zsxkib/step1x-edit) to try it with your own images and prompts.","taskSummary":"Image-to-image is the task of transforming an input image through a variety of possible manipulations and enhancements, such as super-resolution, image inpainting, colorization, and more."},"createdAt":"2025-05-17T19:40:05.000Z","updatedAt":"2025-05-17T19:40:05.000Z"},{"model":"replicate/resnet","classification":{"summary":"Classifies images using ResNet-50 architecture","inputTypes":["image"],"outputTypes":["array"],"task":"image-classification","categories":["computer-vision","image","classification","deep-learning","recognition"],"useCases":["Identify objects in photographs","Categorize product images for e-commerce","Classify wildlife in nature photography","Sort images into content categories","Recognize pet breeds from photos","Filter inappropriate content in image collections","Tag images for searchable databases","Identify landmarks in travel photos","Categorize medical imaging scans","Analyze satellite imagery for land use"],"blogPost":"---\ntitle: ResNet-50: Fast and Accurate Image Classification\nauthors:\n - blog-o-matic\nintro: ResNet-50 delivers reliable image classification for a wide range of applications\npublishedAt: 2023-05-15\n---\n\nResNet-50 is a powerful image classification model that can identify what's in your images with remarkable accuracy. This implementation of the ResNet architecture is designed for efficiency while maintaining high performance across a variety of classification tasks.\n\nThe model accepts an image as input and returns a ranked list of classification labels with confidence scores, making it perfect for applications that need to identify objects, animals, or scenes in images.\n\n[](https://replicate.com/p/64ae6640-b109-4484-9c45-2e6bae918f56)\n\nIn the example above, the model correctly identified a Siamese cat with 88% confidence, followed by Persian cat and tabby as less likely alternatives.\n\nResNet (Residual Network) was developed by Microsoft Research and has become one of the most widely-used architectures in computer vision since its introduction in 2015. The \"50\" refers to the number of layers in this particular implementation.\n\n## Use cases\n\nResNet-50 can be applied to numerous image classification scenarios:\n\n- Automatically tag and organize photo collections\n- Identify products in retail environments\n- Categorize content for moderation systems\n- Recognize animals and plants in nature imagery\n- Sort images for archival and retrieval\n- Classify medical imaging for preliminary screening\n- Identify landmarks in travel photography\n\n## How to use it\n\nYou can run the model with just a few lines of code:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"replicate/resnet\",\n {\n input: {\n image: \"https://example.com/your-image.jpg\"\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model returns an array of the top classifications, each containing the class ID, class name, and confidence score.\n\nTry ResNet-50 today for reliable image classification in your applications!","taskSummary":"Image classification is the task of assigning a label or class to an entire image. Images are expected to have only one class for each image. Image classification models take an image as input and return a prediction about which class the image belongs to."},"createdAt":"2025-05-17T20:12:53.000Z","updatedAt":"2025-05-17T20:12:53.000Z"},{"model":"nohamoamary/image-captioning-with-visual-attention","classification":{"summary":"Generates text captions describing the content of images","inputTypes":["image"],"outputTypes":["text"],"task":"image-to-text","categories":["image","caption","description","content-analysis","accessibility"],"useCases":["Add alt text to website images","Generate social media image captions","Create descriptions for visually impaired users","Catalog image collections with descriptive text","Automate product photo descriptions for e-commerce","Extract content insights from photos","Create image metadata for search engines","Generate captions for video frames","Describe images in educational content","Enable image search with natural language"],"blogPost":"---\ntitle: Generate accurate image captions with visual attention\nauthors:\n - blog-o-matic\nintro: A powerful image captioning model that generates descriptive text for any image\npublishedAt: 2023-10-25\n---\n\nImage captioning bridges the gap between visual content and language, making digital imagery more accessible and searchable. The image-captioning-with-visual-attention model, trained on the Flickr8k dataset, automatically generates natural language descriptions of images by focusing on the most important visual elements.\n\nThis model, created by nohamoamary, uses visual attention mechanisms to generate concise and accurate descriptions of image content.\n\n[](https://replicate.com/nohamoamary/image-captioning-with-visual-attention)\n\nCaption: \"blue car driving over bend in the background\"\n\n[](https://replicate.com/nohamoamary/image-captioning-with-visual-attention)\n\nCaption: \"tan dog is running through field\"\n\n## Use cases\n\nThe model can be applied to numerous scenarios where automatic image description is valuable:\n\n- Creating alt text for images on websites to improve accessibility\n- Generating captions for social media content\n- Cataloging and organizing image libraries\n- Assisting visually impaired users\n- Improving image searchability with text descriptions\n- Content moderation through image understanding\n- Educational applications for visual learning\n\n## Try it yourself\n\nYou can easily run this model with a few lines of code:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\nconst output = await replicate.run(\n \"nohamoamary/image-captioning-with-visual-attention\",\n {\n input: {\n image: \"https://example.com/your-image.jpg\"\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model takes an image URL as input and returns a descriptive caption that captures the key elements of the visual content.","taskSummary":"Image to text models output a text from a given image. Image captioning or optical character recognition can be considered as the most common applications of image to text."},"createdAt":"2025-05-17T20:13:11.000Z","updatedAt":"2025-05-17T20:13:11.000Z"},{"model":"nightmareai/real-esrgan","classification":{"summary":"Upscale and enhance images with optional face correction","inputTypes":["image"],"outputTypes":["image"],"task":"image-to-image","categories":["super-resolution","upscaling","image","restoration","face-enhancement"],"useCases":["Enhance low-resolution photos","Upscale small images","Improve blurry pictures","Restore old family photos","Sharpen digital art","Create printable versions of small images","Fix faces in low-quality photos","Prepare images for large format printing","Enhance social media profile pictures","Improve texture details in photographs"],"blogPost":"---\ntitle: Enhance your images with Real-ESRGAN\nauthors:\n - blog-o-matic\nintro: A powerful image upscaling model with face enhancement capabilities\npublishedAt: 2023-05-15\n---\n\nReal-ESRGAN is an advanced image upscaling solution that produces remarkable results when increasing the resolution of photos, artwork, and other images. This model allows you to scale images by a factor of your choice and even includes an optional face enhancement feature to improve the quality of facial details.\n\n## About Real-ESRGAN\n\nThe [Real-ESRGAN model](https://replicate.com/nightmareai/real-esrgan) is based on the ESRGAN (Enhanced Super-Resolution Generative Adversarial Network) architecture but with improvements that make it more effective at handling real-world image degradation.\n\nWhen you need to enlarge an image without losing quality - or even while gaining quality - Real-ESRGAN is an excellent choice. The model is particularly effective at removing compression artifacts, blur, and noise while preserving and enhancing important details.\n\n## How it works\n\nReal-ESRGAN takes your input image and lets you specify a scaling factor (how much larger you want the output image to be). You can also toggle face enhancement through GFPGAN, which specifically improves the quality of facial features in your images.\n\n## Use cases\n\n- Enhance old family photos\n- Upscale low-resolution images for printing\n- Improve the quality of digital art\n- Restore compressed or degraded images\n- Prepare web graphics for higher resolution displays\n- Enhance facial details in portraits\n- Improve image quality for social media posts\n\n## Try it yourself\n\nYou can easily use Real-ESRGAN in your projects with the JavaScript client:\n\n```javascript\nimport Replicate from \"replicate\";\n\nconst replicate = new Replicate();\n\nconst output = await replicate.run(\n \"nightmareai/real-esrgan\",\n {\n input: {\n image: \"https://example.com/your-image.jpg\",\n scale: 2,\n face_enhance: true\n }\n }\n);\n\nconsole.log(output);\n```\n\nThe model will process your image and return a URL to the enhanced version, which you can then download or display in your application.\n\nReal-ESRGAN is maintained by NightmareAI and provides a reliable solution for image upscaling needs. Whether you're working with old photographs that need restoration or preparing images for large displays, this model delivers impressive results.","taskSummary":"Image-to-image is the task of transforming an input image through a variety of possible manipulations and enhancements, such as super-resolution, image inpainting, colorization, and more."},"createdAt":"2025-05-17T20:13:40.000Z","updatedAt":"2025-05-17T20:13:40.000Z"},{"model":"black-forest-labs/flux-schnell","classification":{"summary":"Fast text-to-image model optimized for local use","inputTypes":["text"],"outputTypes":["image"],"task":"text-to-image","categories":["image","diffusion","prompt-conditioning","fast-inference","local-deployment"],"useCases":["Generate creative concept art quickly","Create custom illustrations for personal projects","Design social media graphics locally","Prototype visual ideas rapidly","Visualize descriptions without internet connection","Produce art for personal websites","Create custom avatars and profile pictures","Generate storyboard frames for videos","Design custom greeting cards","Visualize product designs quickly"],"taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T21:58:16.000Z","updatedAt":"2025-05-17T21:58:16.000Z"},{"model":"lucataco/dreamshaper-xl-turbo","classification":{"summary":"Text-to-image model generating realistic and artistic images","inputTypes":["text"],"outputTypes":["image"],"task":"text-to-image","categories":["diffusion","image","prompt-conditioning","style-transfer","art-generation"],"useCases":["Create photorealistic images from descriptions","Generate fantasy art for games","Design concept art for films","Create digital paintings in various styles","Generate anime and manga illustrations","Visualize architectural or interior design concepts","Create product mockups from descriptions","Generate landscapes and environmental imagery","Design character concepts for creative projects","Create promotional visuals for marketing"],"taskSummary":"Text-to-image is the task of generating images from input text. These pipelines can also be used to modify and edit images based on text prompts."},"createdAt":"2025-05-17T21:59:10.000Z","updatedAt":"2025-05-17T21:59:10.000Z"}]}