Skip to main content

Roadmap

What we shipped, what we're committed to next, and what we're exploring later. Honest about what is on the menu and what is still in the kitchen — three columns, three different levels of certainty.

NowLive

Shipped and live in production

  • sku5 specialty SKUsS1 Background Removal (Fast + HD tiers), S2 Audio Enhancement (denoise / voice-iso / cleanup / master), S3 Speaker Diarization (Brainiall Speaker ID engine), S4 PDF-to-Markdown (Brainiall Document Reader engine), S5 Agent Memory (Brainiall Memory embeddings).
  • skuSpeech AI catalogPronunciation Assessment (Brainiall Pronunciation engine, PCC 0.59 phone — exceeds human inter-annotator agreement). STT in two tiers — Edge (Brainiall Speech (Edge tier, 17 MB), browser/mobile-deployable) and Pro (Brainiall Speech (Pro tier), 99 langs, integrated diarization). TTS (Brainiall Voice (12 voices), top-ranked on independent TTS quality benchmarks).
  • skuNLP SuiteToxicity, Sentiment, Entities, PII detection, Language identification — single Bearer key, OpenAI-compatible.
  • skuImage ProcessingBackground Removal, Upscale (ESRGAN), Face Restore (GFPGAN) — same Bearer key, same response shape.
  • sdkDeveloper artifactsOpenAPI 3.1 spec, TypeScript SDK 1.0.2, Python SDK 1.0.2, Postman collection + environment, hosted Swagger UI at /api-explorer, /quickstart with copy-paste curl/python.
  • complianceCompliance postureGDPR/LGPD compliant by default, signed Data Processing Addendum at /dpa, public Trust Center at /trust, RFC 9116 security.txt for coordinated disclosure.
  • infraDistributionAzure Marketplace SaaS Offer (5 SKUs, certified live), Apify Actor, 8+ MCP registries (Smithery, MCPize, Cline, Cursor, Claude Code Plugin, Apify, etc.), HuggingFace Space.

NextCommitted

Committed for the upcoming quarter

  • docv1.1 head-to-head benchmarksReal measurements vs Krisp / Cleanvoice (S2), AMI 5h corpus (S3), 50-document arxiv eval (S4) — published with raw eval data.
  • infraGPU migration for S1 HD tierBrainiall Cutout engine on CUDA — 10× throughput, ~50% unit-cost reduction. Pricing remains stable (cost passed to margin, not customer).
  • featureBatch endpointsProcess 100s of files in a single request with progress callbacks — 30–50% pricing discount vs per-call.
  • sdknpm + PyPI publishMove from tarball downloads to `npm install @brainiall/sdk` and `pip install brainiall-sdk` with semver-locked releases.
  • complianceSOC 2 Type II auditFormal certification with Big Four auditor — controls already implemented, audit cycle in progress.
  • featureWebhook signing v2HMAC-signed payloads on usage and SaaS Fulfillment events (currently RS256 JWT validated; HMAC adds replay-attack protection symmetric channel).

LaterExploring

Exploring — outcomes drive scope; no fixed timeline

  • complianceCustom data residencyEU-only, US-only, BR-only deployments per tenant — driven by enterprise demand. Architecture supports it; productization depends on signed pipeline.
  • featureEnterprise SSOMicrosoft AD / Okta / Auth0 / Azure AD B2C federation. Currently Bearer-only.
  • featureAudit logs APIPer-API-call provenance feed with retention policy — for regulated workloads that need full traceability.
  • skuPremium specialty tiersS1 batch GPU mode, S3 multi-language diarization, S5 cluster-mode for multi-tenant scale beyond 1M events/month.
  • complianceISO 27001Certification target year 2027 — controls aligned with framework already.
  • featureFunction calling for memoryTool-call interface for S5 Agent Memory — let LLM agents add and query memory via standard OpenAI tool-call schema.

Promises we won't make

Public roadmaps usually leak hype that turns into debt. We avoid that by being explicit about what is intentionally unsaid:

  • Specific dates for "Later" items — exploration depends on outcomes and signed enterprise contracts; fixed dates would be theatre.
  • Feature parity with every competitor — we ship for regulated industries and developer ergonomics, not feature-checklist marketing.
  • Breaking changes without 90-day deprecation notice — even if a Next item replaces a Now item, the old endpoint remains live for at least 90 days.
  • Pricing changes without 30-day notice — active paying subscriptions are price-locked through the notice window; only new sign-ups see the new rate.
  • Re-arrangement of buckets without explanation — when a Next item moves to Now or a Later item is dropped, it shows up explicitly in /changelog.

How we ship

  • Release cadence — platform releases at least weekly (see /changelog); SKU improvements ship continuously and are bundled into the weekly notes.
  • Versioning — paths are versioned (/v1/...); breaking changes land only in a major bump (/v2/...) with deprecation notice and dual-availability for at least 12 months.
  • Subscribe — release notes are emailed to active customers and surfaced on the status page. For real-time platform events, watch status.brainiall.com.
  • Tell us what you'd like to see — feature requests land at /contact or on the public MCP repository GitHub. We read everything; we don't commit to everything.