Compare models by task
Explore the tasks and models you’ll be able to compare side by side. Local comparisons are coming with the upgraded LocalAI Tools.
LLM · Chat & text
Ask questions, draft text, and summarize with large language models (LLMs).
Speech to text (ASR)
Turn spoken audio into text and timestamps.
Streaming speech to text (ASR)
Transcribe incoming audio incrementally for live captions and voice input.
Translation (MT)
Translate text between languages on your device.
Document understanding (OCR)
Extract structured text from document images.
Background removal
Separate a foreground subject from its background.
Text to speech (TTS)
Generate spoken audio from text.
Image upscaling
Increase image resolution locally; the supported scale depends on the checkpoint.
Image understanding
Describe images and ask questions about what they show.
Interactive segmentation
Click a subject to generate a segmentation mask.
Image restoration
Reduce noise or blur, improve document lighting, or straighten pages with task-specific restoration models.
Audio watermarking
Embed a sixteen-bit message in speech audio. Watermarks do not prove authenticity or ownership.
Audio noise removal
Reduce background noise in speech recordings.
Word-level audio alignment
Align transcript words with their positions in an audio recording.
Speaker diarization
Identify speaker turns and build a speaker timeline.
Text embeddings
Convert text into vectors for search and comparison.
Image embeddings & similarity
Represent images as vectors for similarity search and grouping.
Object detection
Find common objects in images and return labeled bounding boxes.
Image classification
Classify images using the checkpoint’s predefined categories.
Image segmentation
Label objects and regions in a scene with pixel masks.
Text reranking
Sort candidate passages by relevance to a query.
Token classification (NER)
Identify and label entities within text.
Depth estimation
Estimate relative depth from a single image.
Pose estimation
Locate body keypoints in images.
Face detection
Locate faces and facial keypoints in images.
Hand landmarks
Locate hands and their landmarks in images.
Audio classification
Explore vocal emotion predictions from audio.
Typed decisions
Ask structured questions about a text state.
Guided image matting
Refine a cutout from an image and a foreground/background/unknown trimap. Returns a soft alpha matte.
Text-guided image selection
Provide an image and a text description. Returns a mask selecting matching pixels.
Text-guided object detection
Provide an image and object descriptions. Returns matching labels, scores, and bounding boxes.
Image & text label matching
Provide an image and candidate text labels. Returns a matching score for each label.
Sentence relationship (NLI)
Provide a premise and a hypothesis. Returns entailment, contradiction, and neutral scores.
Personal information detection
Provide text. Returns labeled spans that may contain personal information; review before redaction.
Image captioning & vision tasks
Provide an image, a selected task, and any required text prompt. Returns task-specific text or regions.
Facial expression classification
Provide a clear face image. Returns visible-expression category scores, not verified emotions.
Text region detection
Locate text regions in an image and return boxes or polygons. Does not read the text.
Document layout detection
Locate headings, paragraphs, tables and illustrations in page images; return regions and reading order.
Face & portrait parsing
Label facial and portrait parts such as skin, hair, eyes and clothing with pixel masks.
Sentiment analysis
Classify the sentiment of text and return category scores.
Toxicity detection
Score potentially toxic text. Predictions need review and depend on the supported language.
Code editing
Provide code and edit context; receive a proposed revised file.
Text continuation
Continue a text prefix with a pretrained base model, rather than an instruction-following chat model.
Table classification
Predict labels for table rows using labeled examples.
Table regression
Predict numbers and quantiles for table rows using examples.
Music beat tracking
Find beats and downbeats in music for rhythm grids and editing.
Audio embeddings
Turn a recording into a numerical vector for sound similarity search and clustering.
Pose-conditioned motion
Generate short motion transitions between skeletal pose conditions; offline animation only.
Optical flow
Estimate the direction and amount of motion between two frames.