\n
![](\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEgit06SKK3XzFZSCJhq4ygask3mob-5PrLcFiDgsT7qXwo5gMmrrlufpNXWev3IZjqESamgtrAtMBsUK4ZVA-Zt9iqXr-th0L8cvQK057_xByadvbJWCzm16yCUBove2WCnhHijUqQHqILY1-aJxfap2BRDK_iPs4egL6J40WeYhQH4fRpnCER3kx-66Ic\/s1600\/O25-BHero-Android-1-Meta.png\")
<\/p>\nPosted by Thomas Ezan \u2013 Developer Relations Engineer, Rebecca Franks \u2013 Developer Relations Engineer, and Avneet Singh \u2013 Product Manager<\/em><\/p>\n We\u2019re bringing back Androidify later this year, this time powered by Google AI, so you can customize your very own Android bot and share your creativity with the world. Today, we\u2019re releasing a new open source demo app for Androidify<\/a> as a great example of how Google is using its Gemini AI models to enhance app experiences.<\/p>\n In this post, we’ll dive into how the Androidify app uses Gemini models and Imagen via the Firebase AI Logic SDK<\/a>, and we’ll provide some insights learned along the way to help you incorporate Gemini and AI into your own projects. Read more about the Androidify demo app<\/a>.<\/p>\n The overall app functions as follows, with various parts of it using Gemini and Firebase along the way:<\/p>\n <\/image><\/p>\n To get started with Androidify, take a photo or choose an image on your device. The app needs to make sure that the image you upload is suitable for creating an avatar.<\/p>\n Gemini 2.5 Flash via Firebase helps with this by verifying that the image contains a person, that the person is in focus, and assessing image safety, including whether the image contains abusive content.<\/p>\n <\/p>\n In the snippet above, we\u2019re leveraging structured output<\/a> capabilities of the model by defining the schema of the response. We\u2019re passing a Schema<\/b><\/span> object via the responseSchema<\/b><\/span> param in the generationConfig<\/b><\/span>.<\/p>\n We want to validate that the image has enough information to generate a nice Android avatar. So we ask the model to return a json object with success = true\/false<\/b><\/span> and an optional error<\/b><\/span> message explaining why the image doesn’t have enough information.<\/p>\n Structured output is a powerful feature enabling a smoother integration of LLMs to your app by controlling the format of their output, similar to an API response.<\/p>\n Once it’s established that the image contains sufficient information to generate an Android avatar, it is captioned using Gemini 2.5 Flash with structured output.<\/p>\n <\/p>\n The other option in the app is to start with a text prompt. You can enter in details about your accessories, hairstyle, and clothing, and let Imagen be a bit more creative.<\/p>\n We\u2019ll use this detailed description of your image to enrich the prompt used for image generation. We\u2019ll add extra details around what we would like to generate and include the bot color selection as part of this too, including the skin tone selected by the user.<\/p>\n <\/p>\n We then call the Imagen model to create the bot. Using this new prompt, we create a model and call generateImages<\/span>:<\/p>\n <\/p>\n And that\u2019s it! The Imagen model generates a bitmap that we can display on the user\u2019s screen.<\/p>\n The Imagen 3 model was finetuned using Low-Rank Adaptation (LoRA)<\/a>. LoRA is a fine-tuning technique designed to reduce the computational burden of training large models. Instead of updating the entire model, LoRA adds smaller, trainable “adapters” that make small changes to the model’s performance. We ran a fine tuning pipeline on the Imagen 3 model generally available with Android bot assets of different color combinations and different assets for enhanced cuteness and fun. We generated text captions for the training images and the image-text pairs were used to finetune the model effectively.<\/p>\n The current sample app<\/a> uses a standard Imagen model, so the results may look a bit different from the visuals in this post. However, the app using the fine-tuned model and a custom version of Firebase AI Logic SDK was demoed at Google I\/O. This app will be released later this year and we are also planning on adding support for fine-tuned models to Firebase AI Logic SDK later in the year.<\/p>\n The app also uses the ML Kit Pose Detection SDK<\/a> to detect a person in the camera view, which triggers the capture button and adds visual indicators.<\/p>\n To do this, we add the SDK to the app, and use PoseDetection.getClient()<\/span>. Then, using the poseDetector<\/span>, we look at the detectedLandmarks<\/span> that are in the streaming image coming from the Camera, and we set the _uiState.detectedPose<\/span> to true if a nose and shoulders are visible:<\/p>\n <\/p>\n The Androidify app makes an extensive use of the Gemini 2.5 Flash to validate the image and generate a detailed description used to generate the image. It also leverages the specifically fine-tuned Imagen 3 model to generate images of Android bots. Gemini and Imagen models are easily integrated into the app via the Firebase AI Logic SDK. In addition, ML Kit Pose Detection SDK controls the capture button, enabling it only when a person is present in front of the camera.<\/p>\n To get started with AI on Android, go to the Gemini<\/a> and Imagen<\/a> documentation for Android.<\/p>\n Explore this announcement and all Google I\/O 2025 updates on io.google<\/a> starting May 22.<\/p>\n<\/div>\n![](\"https:\/\/blogger.googleusercontent.com\/img\/b\/R29vZ2xl\/AVvXsEgGn8cQLt5yYm4cg6RnPqvgDZ2yj-M1Bd-oqsfJAKyfTlDhX3D2FWDzzN0k4NjEMeR0VdUBOc39Sh9WTTNnuHvnqGAkKB45pGw0fr6Gc71UwdhrBrAQSK_xGvNthqD8kHHNDkDFXf4b1l9KzbfJ2FoFEFL3Cii1xUEGH4XDSgI8zeyEb4lKCvi3wIJC4MI\/s1600\/O25-BHero-Android-1.png\")
<\/a><\/p>\nApp flow<\/span><\/h2>\n
<\/div>\nGemini and image validation<\/span><\/h3>\n
val<\/span> jsonSchema = Schema.obj(\n properties = mapOf(\"success\"<\/span> to Schema.boolean(), \"error\"<\/span> to Schema.string()),\n optionalProperties = listOf(\"error\"<\/span>),\n )\n \nval<\/span> generativeModel = Firebase.ai(backend = GenerativeBackend.googleAI())\n .generativeModel(\n modelName = \"gemini-2.5-flash-preview-04-17\"<\/span>,\n \t generationConfig = generationConfig {\n responseMimeType = \"application\/json\"<\/span>\n responseSchema = jsonSchema\n },\n safetySettings = listOf(\n SafetySetting(HarmCategory.HARASSMENT, HarmBlockThreshold.LOW_AND_ABOVE),\n SafetySetting(HarmCategory.HATE_SPEECH, HarmBlockThreshold.LOW_AND_ABOVE),\n SafetySetting(HarmCategory.SEXUALLY_EXPLICIT, HarmBlockThreshold.LOW_AND_ABOVE),\n SafetySetting(HarmCategory.DANGEROUS_CONTENT, HarmBlockThreshold.LOW_AND_ABOVE),\n SafetySetting(HarmCategory.CIVIC_INTEGRITY, HarmBlockThreshold.LOW_AND_ABOVE),\n \t),\n )\n\n val<\/span> response = generativeModel.generateContent(\n content {\n text(\"You are to analyze the provided image and determine if it is acceptable and appropriate based on specific criteria.... (more details see the full sample)\"<\/span>)\n image(image)\n },\n )\n\nval<\/span> jsonResponse = Json.parseToJsonElement(response.text)\nval<\/span> isSuccess = jsonResponse.jsonObject[\"success\"<\/span>]?.jsonPrimitive?.booleanOrNull == true<\/span>\nval<\/span> error = jsonResponse.jsonObject[\"error\"<\/span>]?.jsonPrimitive?.content\n<\/pre>\n<\/div>\nImage captioning with Gemini Flash<\/span><\/h3>\n
val<\/span> jsonSchema = Schema.obj(\n properties = mapOf(\n \"success\"<\/span> to Schema.boolean(),\n \"user_description\"<\/span> to Schema.string(),\n ),\n optionalProperties = listOf(\"user_description\"<\/span>),\n )\nval<\/span> generativeModel = createGenerativeTextModel(jsonSchema)\n\nval<\/span> prompt = \"You are to create a VERY detailed description of the main person in the given image. This description will be translated into a prompt for a generative image model...\"<\/span>\n\nval<\/span> response = generativeModel.generateContent(\ncontent { \n \ttext(prompt) \n \timage(image) \n\t})\n \nval<\/span> jsonResponse = Json.parseToJsonElement(response.text!!) \nval<\/span> isSuccess = jsonResponse.jsonObject[\"success\"<\/span>]?.jsonPrimitive?.booleanOrNull == true<\/span>\n\nval<\/span> userDescription = jsonResponse.jsonObject[\"user_description\"<\/span>]?.jsonPrimitive?.content\n<\/pre>\n<\/div>\nAndroid generation via Imagen<\/span><\/h3>\n
val<\/span> imagenPrompt = \"A 3D rendered cartoonish Android mascot in a photorealistic style, the pose is relaxed and straightforward, facing directly forward [...] The bot looks as follows $userDescription [...]\"<\/span>\n<\/pre>\n<\/div>\n\/\/ we supply our own fine-tuned model here but you can use \"imagen-3.0-generate-002\" <\/span>\nval<\/span> generativeModel = Firebase.ai(backend = GenerativeBackend.googleAI()).imagenModel(\n \"imagen-3.0-generate-002\"<\/span>,\n safetySettings =\n ImagenSafetySettings(\n ImagenSafetyFilterLevel.BLOCK_LOW_AND_ABOVE,\n personFilterLevel = ImagenPersonFilterLevel.ALLOW_ALL,\n ),\n)\n\nval<\/span> response = generativeModel.generateImages(imagenPrompt)\n\nval<\/span> image = response.images.first().asBitmap()\n<\/pre>\n<\/div>\nFinetuning the Imagen model<\/span><\/h3>\n
<\/div>\nML Kit<\/span><\/h2>\n
private<\/span> suspend fun<\/span> runPoseDetection<\/span>() {\n PoseDetection.getClient(\n PoseDetectorOptions.Builder()\n .setDetectorMode(PoseDetectorOptions.STREAM_MODE)\n .build(),\n ).use { poseDetector ->\n \/\/ Since image analysis is processed by ML Kit asynchronously in its own thread pool,<\/span>\n \/\/ we can run this directly from the calling coroutine scope instead of pushing this<\/span>\n \/\/ work to a background dispatcher.<\/span>\n cameraImageAnalysisUseCase.analyze { imageProxy ->\n imageProxy.image?.let { image ->\n val<\/span> poseDetected = poseDetector.detectPersonInFrame(image, imageProxy.imageInfo)\n _uiState.update { it.copy(detectedPose = poseDetected) }\n }\n }\n }\n}\n\nprivate<\/span> suspend fun<\/span> PoseDetector.detectPersonInFrame(\n image: Image,\n imageInfo: ImageInfo,\n): Boolean {\n val<\/span> results = process(InputImage.fromMediaImage(image, imageInfo.rotationDegrees)).await()\n val<\/span> landmarkResults = results.allPoseLandmarks\n val<\/span> detectedLandmarks = mutableListOf<\/div>\nGet started with AI on Android<\/span><\/h2>\n