Reading + recording development build v0.1.0
All guides

Build status

Features and current limits

Check what is implemented, where the limits are and how AI services process data.

Capability overview

Features and limits in the development build

Reading, translation, notes, document questions and recording tools, with their implementation status and current limits. AI teamwork remains planned.

  1. Reading and research 10 Features

    Read documents in context, translate into Traditional Chinese for Taiwan, keep source-linked notes and check answers against the original.

    • Import PDF, Word (.docx), Markdown and UTF-8 text — Working today

      PDF up to 200 MiB, Word up to 100 MiB, Markdown and text up to 20 MiB. Older .doc files are not supported.

    • Switch between Original, Translation and Side by side — Working today
    • PDF layout translation, current-page translation and saved progress — Limited today

      Translated text overlays the original text areas and may shrink or clip on dense pages. Full-document and current-page translation can be stopped and resumed. Exporting a translated PDF is not implemented.

    • Translate Word, Markdown and text as flowing passages — Working today

      The current target is Traditional Chinese for Taiwan. Translation requires the built-in local model; original Word pagination is not preserved.

    • Search original and translated passages, follow an outline and focus on reading — Working today
    • Highlight passages, add comments and quote sources into Markdown notes — Working today
    • Generate Study notes with section key points and passage links — Working today

      Requires the built-in local model. Generated points need checking; incomplete sections can be retried.

    • Ask about this document or all documents, with cited sources — Working today

      Uses the selected provider. A remote selection sends the question and retrieved passages to that service. Earlier turns are not passed as conversational context.

    • Open PDF links and recognized numbered references — Limited today

      Supports known bracketed numeric references and supported PDF links. It does not automatically parse all author–year or arbitrary citation styles, or retrieve the referenced paper.

    • Read scanned PDFs and write your own notes — Limited today

      OCR is not available yet. Translation, highlights, Study notes and document questions need extractable text.

  2. Recording and transcription 7 Features

    Records in-room and online meetings to your own disk and turns them into text, with no remote service involved.

    • Captures the room and the online call together — Working today
    • Every recording saved automatically on your own disk — Working today
    • Local speech recognition with language selection — Working today
    • Import earlier recordings and transcribe them — Working today
    • Import YouTube audio on macOS and Windows — Limited today

      Available in the development build on macOS and Windows.

    • Recovery of committed recording work after interruption — Limited today

      Uncommitted audio or text can be missing. Current installed crash and power-loss verification remains open.

    • Speaker labels, for now your side and the call's side — Limited today

      Your microphone is labelled You and the call's audio Others. Telling several people in one room apart is planned, not built.

  3. The meeting workspace 7 Features

    Stay focused during the meeting, then review quickly and draft the follow-up afterwards.

    • Notes and transcript side by side, for context at a glance — Working today
    • A visual timeline to find key moments quickly — Working today
    • Click any line to play the recording from that moment — Working today
    • Playback speed, ten-second skips and audio reload that keeps your place — Working today
    • Follow-up email and status-update drafts, and flashcards of the key points, in one click — Working today
    • Full keyboard shortcuts — Working today
    • English and Traditional Chinese interface; some template names remain in English — Working today
  4. Summaries and action items 6 Features

    Turn discussion into decisions and action items. The built-in model can summarize after recording; selected external providers are started manually. Each summary identifies where it was processed and can be regenerated.

    • Generate or regenerate a summary by hand at any time — Working today
    • Custom summary templates that match your company's format — Working today
    • Choose the language a summary is written in — Working today
    • Concrete action items and owners, extracted automatically — Working today
    • Automatic post-recording summaries and action items with the built-in model — Working today
    • A live summary beside the transcript while you record — Limited today

      The Studio development build offers live summaries on macOS. Windows requires hardware, runtime and model checks; 8 GB of video memory and 16 GB of system memory are provisional thresholds, not measured minimums or performance guarantees. Without live summaries, supported configurations can summarize after recording.

  5. Search, Ask and recall 5 Features

    Months later, find the decision and the exact words, so past discussions become knowledge your team can look up.

    • Search every meeting's decisions at once — Working today
    • Answers to your questions, grounded in past meetings — Working today
    • Indexed automatically the moment a recording ends, ready to search — Working today
    • Recall open items from recurring meetings; title matching currently uses Latin letters and digits only — Working today
    • A knowledge map of what each meeting covered, and how it connects to the rest of your library — Working today
  6. Models and engines 5 Features

    You choose which models do the work on your device, weighing privacy, speed and cost.

    • A built-in local AI engine, with no separate server to set up — Working today
    • Model downloads verified, so files arrive complete and untampered — Working today
    • Connect to a model server your company hosts — Working today
    • Custom vocabulary that helps speech engines get your names, terms and acronyms right — Working today
    • Re-transcribe a finished recording with another on-device engine — Working today
  7. Your data, and getting it out 8 Features

    Stored in open formats, with Studio’s network purposes and independent process routes listed so you can understand and take your data with you.

    • Stored on your disk in open formats — Working today
    • Export to common formats to share with your team — Working today
    • Date-based usage reports with CSV and PDF exports — Working today
    • Deleting a meeting removes its recordings, records and index entries — Working today
    • A database integrity check at startup, with a guided repair that never overwrites your files — Working today
    • Connect a calendar subscription link for meeting reminders — Working today
    • Read summaries and answers aloud — Limited today

      Windows only, using the voices built into Windows.

    • Secure in-app updates — Limited today

      The update mechanism is built and will switch on once public installers ship.

“Working today” means implemented in the Studio development build; no signed installer has passed native release QA yet. Document translation and Study notes require the built-in local engine. Automatic meeting summaries, knowledge maps and live summaries also use that engine. Manually selected external providers can process meeting summaries or Ask; document Ask sends the question and retrieved passages to the selected provider.

See today's screens

Studio and AI services

Development-build and service questions.

How reading and recording in the development build use your data, models and AI services, with future AI teamwork identified separately.

Back to the guide

Data and AI services5
Will Studio upload my meeting content?

Studio development builds default to local recording, transcription, summaries and questions. Selecting external AI sends the material needed for the request to that service. OptiVerse Studio also plans to use external AI services; the destination depends on how the work is processed.

Do I need an account or API key?

Studio’s local recording tools do not need an AI service account, but speech and summary models must be downloaded first. External AI may require a sign-in or API key. OptiVerse Studio service connections are still in development.

Where does Studio store my data, and can I delete it?

Recordings, transcripts, summaries, imported documents, translations, notes and local indexes are stored on your computer. Deleting a meeting or document removes the related data managed by Studio. Exported files, backups and content already received by external providers must be managed separately.

Where does the OptiVerse Studio AI team work?

Codex with a ChatGPT subscription and Claude Code are planned execution routes; the full connections still need validation. These services receive the material you provide to the AI and require an internet connection.

Can Studio use our company’s AI server?

Meeting summaries and meeting or document Ask can use Ollama on this computer, an HTTPS company server or supported cloud services. Relevant content goes to the provider you select; API keys stay in the operating system’s credential store. Document translation and Study notes require the built-in local model.

Computers and languages3
Which computers are supported?

Studio development builds target Macs with Apple silicon or Intel processors and 64-bit Windows computers with AVX2. Live features depend on hardware and models; see “Before you start” in the setup guide for requirements. The future AI team workflow is Mac-first, with Windows support still to be confirmed. Check computer requirements

Which features work offline?

With models downloaded, local recording, transcription, built-in summaries, document translation, Study notes and search can work offline. Ask depends on the selected provider. Model downloads, update checks, YouTube imports and external AI services need the internet.

Which languages are supported?

The development build has English and Traditional Chinese interfaces. Speech recognition is tuned for Taiwanese Mandarin and English, and meeting summary language can be chosen. Document translation currently targets Traditional Chinese for Taiwan. The planned AI team version will prioritize Taiwanese Traditional Chinese, with English later.

Reading and research4
Which documents can I read and translate?

Import PDF, Word (.docx), Markdown and UTF-8 text files. Scanned PDFs without a text layer can be read and used with your own notes, but OCR, translation, highlights, Study notes and Ask are not available for those scans.

Does translation preserve every page exactly?

PDF translation places text in the original page areas; dense passages may shrink or clip with a warning. Word, Markdown and text use flowing passages. The target language is currently Traditional Chinese for Taiwan, and translated-PDF export is not implemented.

Are Study notes the same as a chat or summary?

Study notes collect section key points with passage links. Document Ask answers one current question using retrieved passages from this document or all documents; previous turns are not sent as conversational context. Continuing AI team discussion remains planned.

Do translation and document questions use the same model?

Translation and Study notes require the built-in local model. Document Ask uses the selected provider; if that provider is external, your question and retrieved source passages are sent to it. Download local models before using their offline features.

Studio recording tools2
Can it identify different speakers?

Studio currently distinguishes audio sources: the microphone is labelled “You” and call audio “Others”. Automatic identification of multiple speakers within the same audio source is planned.

How accurate are transcripts and summaries?

Recording quality, accents, specialist terms and the selected model all affect results. Check important points against the recording and cited passages; generated summaries may also need correction.

About this guide

These guides cover reading and recording in development build v0.1.0, which has no public installer yet. Implemented features still need release validation. AI teamwork remains planned.

Planned AI team workflow

A question this guide does not answer?

Write to us, and attach the log file if something went wrong. Logs hold no meeting content.

Email us