Tinia · An AI + node-driven acoustic insight engine
No-code, drag-and-drop
Empowering every engineer to build a professional acoustics analysis system
Pro Nodes
37+ Official Acoustic Nodes
Analysis Flow
Drag to build, runs in seconds
Technical Barrier
80% lower with no-code
Learning Curve
Productive on your own in 1-2 weeks
Deployment
Desktop · works offline
AI Capability
Conversational build + MCP collaboration
01 — Product Story
Watch the patterns repeat, let the hidden order emerge
Every dataset carries a story of its own. Tinia believes the essence of analysis is not computation but listening — build the flow, and let the hidden patterns surface on their own, so that every analytical decision is visible, reproducible, and inheritable.
The Name
Tinia
Taken from the supreme deity of Etruscan mythology, who presided over insight, order, and boundaries — he saw everything yet never intervened, simply letting the patterns reveal themselves. The essence of data analysis is no different:
Not imposing meaning on the data, but building the flow so the hidden patterns surface on their own.
Product family · bestfunc Intelligent Industrial Suite
02 — Philosophy
Six principles that redefine industrial acoustic analysis
Tinia does not bolt AI features onto a traditional tool. It redesigns the operating system for industrial acoustic analysis — with AI as a first-class citizen, a node ecosystem as the growth engine, and DAG flows as the carrier of knowledge. These six principles are the core product philosophy that sets Tinia apart from conventional NVH tools.
Visual Flow
Nodes are logic, links are reasoning, the flow is the document
Every analysis capability is encapsulated as an independent node; nodes connect into a DAG (directed acyclic graph) that the engine schedules automatically by topology. Not script code, not screenshot tutorials — the flow is a knowledge asset: readable, editable, reusable, and directly orchestratable by AI.
Versus traditional: scripted and imperative vs. DAG declarative; logic buried in code vs. a visible analysis chain; changes by guesswork vs. parameter tweaks previewed instantly.
Plugin Node Ecosystem
Build your own nodes, deploy from a private repository, plug in algorithms instantly
DevStudio in-browser IDE + Python SDK + TSX front-end framework. Developers write algorithms, configure the UI, and publish to the store; users subscribe and install in one click. Organization-private node pools are supported, so core algorithms are wrapped as nodes and intellectual property never leaves the company.
Moat: HEAD/Testlab spent 30 years developing everything in-house; Tinia's node store lets the whole industry contribute, so the algorithm ecosystem keeps expanding over time — an order-of-magnitude gap.
AI-Assisted Build
Describe your intent, let AI orchestrate the nodes, lower the barrier to entry
A built-in MCP server makes AI a first-class client. Engineers describe what they need in natural language, and AI selects nodes, sets parameters, wires the flow, and runs the analysis. Iterate through ongoing dialogue — from 30 minutes of manual building to 10 seconds of AI generation.
Architectural advantage: not an AI button bolted on later — it was designed for AI workflows from v1.0. Adding MCP to conventional desktop software means rewriting authentication and the event bus: 3-5 years of engineering.
Reproducible Decisions
Every analysis step traceable, auditable, and retained
Every run creates a GraphRun instance that records node states, input/output blob handles, start and end times, and error tracebacks. Content-addressed blobs guarantee the same input → the same result. AutoML distills discriminant functions into readable formulas — the black box becomes a white box.
Compliance value: from raw data → analysis flow → metrics → decision → conclusion, every step is clickable and traceable. Meets the data-integrity requirements of ISO 17025 laboratory accreditation.
Real-Time Insight
Spectrum, waveform, metrics — instant node previews, analysis is WYSIWYG
Every node ships with its own visual Viewer — spectrum plots, bar charts, heatmaps, metric tables — with limit-line overlays and band-step rendering. After a parameter change only the affected nodes re-run, updating in seconds. Dashboards combine multiple Viewers freely and can be saved and shared.
Efficiency gain: eliminates the fragmented "run analysis → export data → switch tools to plot" experience. Debugging is 95% more efficient than changing parameters and re-running the whole flow in conventional tools.
Multi-User Collaboration
Organization-level data isolation, group-based permissions, shared analysis assets
A multi-tenant Org system isolates data sources, flows, nodes, and dashboards at the organization level. Seat management plus role-based permission assignment. Flow template libraries, data sources, and Dashboards are shared across the team. Full audit logs of AI operations meet enterprise compliance requirements.
Team efficiency: build an organization-level template once and the whole team reuses it. Enterprise-private node pools keep core algorithms inside the company. Knowledge turns from individual skill into an organizational asset.
CLOSED LOOP The six principles form a complete closed loop: Visual Flow makes analysis logic transparent and readable → the Plugin Ecosystem expands capability without limit → AI-Assisted Build lets everyone get started → Reproducible Decisions make results trustworthy and auditable → Real-Time Insight makes analysis WYSIWYG → Multi-User Collaboration turns personal knowledge into an organizational asset. Stacked together, these six layers move us from "tool replacement" toward a new paradigm of "AI engineer + node store".
03 — AI NATIVE
Talk to it like a colleague, build your analysis flow
Tinia's AI is not a chatbot but an intelligent assistant that operates the platform directly. Built on the MCP (Model Context Protocol), an AI client can create flows, add nodes, configure parameters, connect ports, run analyses, and view results — fully automated end to end.
What AI Can Do
Natural-language intent understanding
"Show me the spectrum of this e-drive noise recording and how it evolves between 1000 and 3000 RPM, and label the main orders"
Auto-select nodes, set parameters, wire the flow
AI turns your request into a complete analysis pipeline — from intent to a runnable flow in 10 seconds
Iterate through ongoing dialogue
"Change the filter cutoff to 100Hz", "add loudness analysis" — natural-language edits that take effect instantly
Result interpretation and recommendations
"What causes the 1200Hz peak in the spectrum?" — AI analyzes the result and offers expert interpretation plus next-step advice
DAI Collaboration (MCP) Authorize AI clients in "Account Settings", follow / ask back
AI client authorization (Claude Code, etc.)
MCP OAuth + scope-isolated permissions (code / data / data write)
Shipped
MCP tool set (20+)
Four tool groups — flow / dev / datasource / plugin — let AI fully automate flow edits and node authoring
Shipped
Global AI activity panel (follow / observe)
The right-hand drawer shows what AI is doing in real time + a project lock halo
Shipped
AI bounce-back questions (elicitation)
When AI is unsure, an interrupt bar appears and it continues after the user chooses
Shipped
DDeveloper Tools (projects / nodes / live preview) The "Developer Tools" page — create a project / pick a template / write code / preview
Project management
Create / edit / delete / share dev projects with organization members
Plugin store: browse / my subscriptions / my submissions / pending review
Sidebar "Plugins" + top "Store" menu — publish / review / subscribe to free plugins
Shipped
Paid plugin distribution channel
Opens in phase two (revenue-share model)
Phase Two
ECOSYSTEM From "tool user" to "platform builder": algorithm engineers wrap core algorithms into nodes with the Python SDK → publish them to the enterprise-private store → application engineers subscribe to nodes and build templates → QC staff run inspections with those templates → everyone gets up to speed through AI dialogue. Knowledge turns from individual skill into enterprise asset.
04 — Data Sources
Data Sources · 10 Items
Sidebar menu "Data Sources". The datasource detail page has 3 tabs: overview / files / channel naming.
Merge multiple single channels into a multichannel set, 3 pairing modes — the "Compose" button on the detail page, status: composed / pending composition
Shipped
Channel naming templates (save / apply)
Continuously refined from v1.23 to v1.31
Shipped
Channel calibration (dB offset)
Filled in the channel naming template; downstream level meter / psychoacoustic nodes apply it automatically
Shipped
Channel select / channel split nodes
channel_select / channel_split — used inside flows
Shipped
The "Files" tab (search / select all / bulk delete / bulk download)
The datasource detail page UploadManager — view / delete / bulk-archive and download files
Shipped
Max size per datasource
C ≤ 5 GB / unlimited on Pro
Planned P1
Max items per flow
C ≤ 100,000 / unlimited on Pro — large datasets are the Pro selling point
Planned P1
Max number of datasources
C ≤ 50 / unlimited on Pro
Planned P2
05 — Nodes
37 official nodes + plugin ecosystem · 8 categories
All 37 official acoustic nodes are fully open to the Community edition — this is Tinia's foundational product commitment.
ShippedAvailable in v1.32PlannedNot implemented
NPreprocessing Audio split / active segment / channel select / channel split / filtering / weighting — 9 in all
Audio split · active segment detection
audio_segment_split / active_segment — split by time window / silence / equal length
Shipped
Channel select · channel split
channel_select / channel_split — pick from multichannel input / split into single channels
score_predictor — apply the discriminant function to new data to produce a score
Shipped
Official free plugins · community free plugins
Annotation system v1 integration + the free section of store.bestfunc.com
Shipped
Commercial plugins (paid)
Exclusive to Tinia Pro; opens once plugin-protection Phase B (Cython compilation) ships
Phase Two
VIEWER Every node ships with visual output — spectrum plots, bar charts, scatter plots, heatmaps, box plots and more — so results are clear at a glance. Parameter changes update in seconds, WYSIWYG. Supports limit-line overlays, band-step rendering, and other professional NVH visualization styles.
06 — AutoML Jobs
AutoML Jobs · auto-tuning + discriminant-function distillation
Sidebar menu "AutoML Jobs", entered via the "Auto-Tuning" button at the top of the flow editor. 4-step setup + trial list + detail page (parameter importance / Top-K) + diagnostics modal (3 tabs).
M4-step setup wizard Entered via the "Auto-Tuning" button; runs after the canvas is locked
Step 1: choose a datasource
Pick from the dataset_node candidates on the canvas
Shipped
Step 2: choose the grouping dimension
Group by datasource attribute / item property (OK/NG, etc.)
Shipped
Step 3: choose tuning / evaluation
Choose the search_space (parameters flagged tunable on the node) + the evaluation unit
Shipped
Step 4: configure + submit
Number of trials / search algorithm / stopping criteria; runs after forking a snapshot
Predicted value / actual / misclassified-in-red for every item
Shipped
MThe "Discriminant Function" tab (model distillation) Click "Generate Discriminant Function" → distillation across 5 algorithms + KaTeX formula rendering
Generate discriminant function (5-algorithm distillation)
LR / LDA / polynomial / decision tree / GBDT — approximating the black box with explainable models
Shipped
Formula rendering (KaTeX)
Renders the distilled LR / LDA formulas as readable LaTeX
Shipped
"Create Scoring Node" (one-click closed loop)
Forks a snapshot from the discriminant function + automatically appends AutoML score prediction + chart viewer
Shipped
Score list + scatter plot + linkage
Score ranking per item + clicking a table row highlights the scatter point
Shipped
Export diagnostics JSON
Includes the distillation result for external use
Shipped
MItem Preview Click a single sample in the diagnostics view → an audio player + spectrum plot appear
Click a table row / scatter point → item preview
Links the ScoreRanking table row + scatter point
Shipped
Item preview panel: audio player + spectrum plot
Pulls the spectrum from the evaluation node + the audio uri from the datasource
Shipped
AUTOML The core value of AutoML: from hand-probing parameters to systematically searching for the optimal configuration. Built on the Optuna engine and supporting four search strategies — Bayesian optimization (TPE) / random / grid / CMA-ES — it finds the best parameter combination automatically. When the job finishes, one click distills an explainable discriminant function (LR / LDA / decision tree / GBDT), renders the formula with KaTeX, and generates a scoring node to close the loop.
07 — Scenarios
Covering six typical industrial application scenarios
Tinia already serves customers in automotive components, home appliance manufacturing, industrial equipment, and new energy, covering the full range of needs from R&D validation to production-line quality inspection.
Production-line abnormal noise detection
Online and offline abnormal noise detection for motors, compressors, pumps, and other rotating machinery, covering typical fault modes such as squeal, knocking, and rubbing.
FFT spectrum + octave analysis → frequency-domain features
Template reuse → fast rollout across multiple product lines
Automotive NVH analysis
NVH performance evaluation for complete vehicles and components: order tracking, transfer path analysis, and subjective sound-quality metric calculation.
Order tracking → extraction of rotational-order components
Campbell diagram → 3D order × RPM visualization
Transfer function FRF → structural path contribution
TNR + tonality → squeal / pure-tone detection
Predictive maintenance (PdM)
Vibration condition monitoring and fault warning for bearings, gearboxes, fans, and other critical equipment, with ISO-standard grading.
Metric comparison reports → quantified evaluation of multiple designs
Teaching and research
Lab teaching for acoustics and vibration courses, graduate research topics, and rapid prototyping of algorithms. Tinia Community is free forever.
Signal generator → synthesize test signals with known characteristics
All nodes free and open
Up and running in 1-2 weeks, no programming background required
AI conversational guidance eases the teaching load
QC data management and reporting
Establish standardized analysis flows, auto-generate QC reports, keep data traceable and results comparable.
Dashboard editor → custom data walls
Template version control → flows that iterate and roll back
Report export → PDF / CSV / PNG
AI result interpretation → automatically generated conclusions and recommendations
08 — Developers
An open platform, infinitely extensible
Tinia is not a closed system. The Python SDK + TSX front-end framework + MCP AI collaboration protocol let enterprises and third-party developers extend the platform on their own terms and accumulate core algorithm assets.
4 project templates
Basic node / analysis node / datasource plugin / empty project — scaffolded in one click
Python hot reload
Saving a .py file reloads the node automatically, no service restart
Live TSX view compilation
Bundled with esbuild: write parameter forms / Viewers / help pages, saved is live
Browse / subscribe / publish / review, with support for enterprise-private and official stores
MCP AI collaboration
20+ tools let AI clients fully automate flow edits, node authoring, and analysis runs
Version control
dev project-level snapshots, switching, restore, and diff — code changes fully traceable
tinia CLI
Connects local repositories to the desktop app via the tinia:// scheme and OAuth authorization
FROM USER TO BUILDER From "tool user" to "platform builder": algorithm engineers wrap core algorithms into nodes with the Python SDK → publish them to the enterprise-private store → application engineers subscribe to nodes and build templates → QC staff run inspections with those templates → everyone gets up to speed through AI dialogue. Knowledge turns from individual skill into enterprise asset.
09 — Get Started
Three steps to your first analysis
①
Download and install
Tinia Desktop — double-click to launch, all dependencies bundled, no environment setup
②
Import data
Drag and drop WAV and other audio formats, or connect to enterprise datasources
③
Start analyzing
Drag nodes, draw links, hit run — or just tell the AI what you need
→
Keep advancing
Save templates, build custom nodes, compose dashboards, and construct an enterprise-grade analysis system
All 37 official acoustic nodes, the full flow editor, visual viewers, and developer tools are open. Download now — in 1-2 weeks you can run professional acoustic analysis on your own.
Tinia — the next-generation industrial acoustics platform
No-code · drag-and-drop · 37+ nodes · AI conversational build
