Local processing
AI Training Dataset Converter
Map single-turn CSV or JSON question and answer fields into messages or Alpaca JSONL, with an explicit row report.
Open toolA clearer path through your AI workflow
Free Local AI Tools: OCR, Word, Datasets and RAG
Recognize image text, export Markdown to Word, review private prompt values, prepare JSONL and Batch files, estimate LLM memory and evaluate RAG in your browser.
Browser-local toolsNo API keyEnglish & 中文
Ready to use
Your AI workbench
Start with an example, adjust the parameters and download a real result. Each tool explains its method and what the result can tell you.
16 tools shown
Local processing
AI Dataset Train / Validation / Test Splitter
Split JSONL reproducibly with a seed, explicit ratios, optional duplicate removal and a field that keeps groups together.
Open toolLocal processing
AI Response JSON Extractor
Find complete JSON objects and arrays inside mixed response text, select a candidate and export without rounding numeric values.
Open toolLocal processing
AI Response Text Cleaner
Apply explicit outer fence, whitespace and newline rules to existing AI text while comparing the original and cleaned result.
Open toolLocal processing
AI Embedding Similarity Calculator
Compare existing vectors with cosine, angle, dot product, L1/L2 distance, coordinate contributions and candidate rankings.
Open toolLocal processing
RAG Retrieval Evaluator
Evaluate existing ranked document IDs against graded judgments with Precision, Recall, F1, Hit, RR, AP and nDCG at K.
Open toolLocal processing
LLM Sampling Playground
Explore temperature, min-p, top-k, top-p and greedy selection on your own logits with probability bars and repeatable seeded sampling.
Open toolLocal processing
AI PNG Metadata Viewer
Read embedded PNG text locally, inspect Automatic1111 parameters and ComfyUI prompt or workflow JSON, and export the original fields.
Open toolLocal processing
Receipt Photo to CSV
Recognize a receipt photo in your browser, review the merchant, date and total, and export a CSV expense row.
Open toolLocal processing
AI Old Photo Restoration & Colorization
Restore scratched or faded black-and-white photos and run the Alibaba DAMO DDColor neural network directly in your browser. Preserve 100% original resolution via CIE Lab luminance locking with 7 film presets and 1:1 split comparison.
Open toolLocal processing
Image to Text OCR
Recognize English or Simplified Chinese text from a local image, with a preview, preprocessing, progress and editable output.
Open toolLocal processing
AI Markdown to Word Converter
Turn an existing Markdown answer into editable Word, with paper, font, spacing, margins and a safe content preview.
Open toolLocal processing
AI Prompt Privacy Redactor
Review fixed-pattern matches and custom literal words, then replace selected values before sharing a prompt.
Open toolLocal processing
AI JSONL Dataset Deduplicator
Find duplicate raw rows or matching text fields with explicit normalization, conflict review and preserved originals.
Open toolLocal processing
LLM Weight and KV Cache Memory Calculator
Estimate theoretical model weight and full-attention KV memory from explicit parameters and your overhead budget.
Open toolLocal processing
OpenAI Batch JSONL Builder and Result Matcher
Prepare text-only Batch request files and align existing result lines by custom_id without an API key.
Open toolKeep going
Pair it with your Agent toolkit
Token Counter & Vocabulary ComparisonCount exact o200k_base or cl100k_base text tokens locally, inspect token IDs and compare both vocabularies.Agent Context Budget PlannerAllocate a user-defined context window across editable text segments, output reserve and extra overhead.RAG Text Splitter with Source OffsetsPrepare traceable text chunks with paragraph, line or fixed boundaries, a hard size cap and adjustable overlap.JSON Schema BuilderInfer a draft 2020-12 JSON Schema from sample JSON and validate another value locally.Agent Output Evaluation RunnerCheck existing model outputs with deterministic assertions, Unicode lengths and normalization settings.Agent Cost EstimatorEstimate token charges from your own rates, tokens per step, step count and repeated runs.
Official external links
When you need a model to run
Use the official destination for generation, source-based research or local inference. These services run separately from UTools; their account, hardware, privacy and billing rules apply.
6 / 6 resources
Image workflows
ComfyUI
Build node-based image workflows and reopen workflows embedded in compatible output files.
- Before you start
- Local installation needs models and suitable hardware. Hosted options have their own accounts, limits and charges.
- Where data goes
- Local nodes and connected API nodes have different data paths. Inspect a workflow and its custom nodes before running it.
Multimodal playground
Google AI Studio
Explore supported Gemini, media, speech and Live models in Google’s playground.
- Before you start
- Google account and eligible region; available models, quotas and billing depend on your account and selected service.
- Where data goes
- Inputs are sent to Google when you use the service. Check the terms for your account before submitting private material.
Research with sources
Gemini Notebook / NotebookLM
Ask questions about uploaded sources. Current Google help uses the Gemini Notebook name.
- Before you start
- Eligible Google account, age and region. Work and school access may require an administrator to enable the service.
- Where data goes
- Sources are uploaded to Google. Sharing settings and account-specific terms determine access and handling.
Explore model demos
Hugging Face Spaces
Find interactive machine learning demos published by individual developers and organizations.
- Before you start
- Each Space has its own availability, hardware, queue and access rules. Creating compute-backed apps may require a paid plan.
- Where data goes
- A platform listing is not an endorsement of each app. Check its publisher, code and data handling before uploading files.
Run models locally
Ollama
Download compatible models and run inference on your computer; cloud models are a separate option.
- Before you start
- Install the runtime, download model weights and check memory, storage and hardware support.
- Where data goes
- Local and cloud modes differ. Official docs describe a local-only setting that disables cloud models and web search.
Run models locally
LM Studio
Run downloaded compatible language models with a desktop interface and local document chat.
- Before you start
- Install the app and download supported model files first. Model size and hardware determine whether it fits your computer.
- Where data goes
- Offline local operation is documented. Check the selected model connection and any remote features before handling private files.
Public documentation checked · No sign-in, installation or service execution was tested. Resources are grouped by task, without traffic, rating or ranking claims. Check current platform rules before use.
Frequently asked questions
What works here, and where to go next.
Do these tools generate answers with a model?
The browser tools prepare files, review privacy, convert Markdown and calculate explicit metrics. Image OCR and receipt recognition run a local Tesseract model in a Worker. Other tools do not generate model answers. Cloud model platforms are separately labelled official links.
Do I need an account or API key?
The browser tools are free and need neither an account nor an API key. External platforms have their own access rules, limits, hardware needs and charges.
Where are my files and prompts processed?
On-site tools process inputs in your browser without storing their contents. Page-view analytics do not contain tool input. When you open an external service, review that service’s data rules before uploading anything.