Audio Memory

Hana Audio Memory lets organizations transform spoken content into searchable knowledge. Upload an audio or video file, and Hana transcribes the speech into text, optionally preserving segment‑level timestamps so you can reference or revisit exact moments later. This page explains how the feature works, who can use it, and what to expect throughout the ingestion process.

---

Overview

• Purpose: Convert recorded speech (meetings, interviews, voice notes, etc.) into “memories” that Hana can retrieve during chats or other workflows.
• Output: The original file is processed to produce a transcript. Only text (plus optional timestamps) is stored; the raw audio is discarded.
• Storage: Each upload becomes a dedicated memory batch, appearing alongside other batches in the organization’s memory management area.

---

Access & Requirements

Requirement	Details
Role	Only Admins and above can create audio memories within their organization. Super Admins can upload on behalf of any user.
Plan	Available on Basic plans and higher. Free‑tier organizations cannot ingest audio memories.
Quota	Each batch counts against the organization’s memory quota, just like other memory types.

---

Supported Formats & Limits

• Maximum size: 25 MB per file. Larger uploads are rejected.
• Accepted types:
  • Audio: mp3, mpeg, mp4, m4a, mpga, webm, wav
• Rejection triggers: Unsupported format, oversized file, or antivirus failure.

---

Transcription Models

Hana automatically chooses a transcription model based on plan tier:

Plan	Model
Pro and above	High‑accuracy
Basic	Cost‑efficient
Any plan with timestamps enabled	Uses a model that supports segment‑level timing for more detailed results.

---

Timestamp Option

When uploading, you may enable Include Timestamps:

• Each transcript segment is tagged with a start time and duration in seconds.
• These markers help Hana reference or quote specific moments in later conversations.
• Expect slightly longer processing times and larger transcript output when this option is enabled.

---

Upload Workflow

1. Start an Audio Memory Batch

   • Navigate to the memory management area and select the audio upload option.
   • Super Admins can also initiate uploads on behalf of users by specifying a user ID.

2. Select the File

   • Choose a supported audio or video file up to 25 MB.
   • Optional: toggle Include Timestamps to retain segment‑level timing.

3. Security Scan

   • Hana scans the file for malware. Infected files are immediately rejected.

4. Transcription

   • The platform transcribes the speech with the appropriate model.
   • If timestamping is enabled, segments are annotated with start and duration.

5. Batch Creation & Progress

   • A new memory batch is created and queued for ingestion.
   • Progress indicators show statuses such as In Progress, Ingested, Partially Ingested, Error, or Aborted.

6. Completion

   • Once ingestion is finished, the transcript is stored as a searchable memory.
   • The batch appears in memory listings, where it can be edited, resynced, or deleted like any other batch.

---

Using the Transcript

• Search & Retrieval: Audio transcripts are indexed with other memories. Hana can reference them when answering questions or generating content.
• Referencing Segments: If timestamps were captured, you can ask Hana about specific moments (e.g., “What was discussed at 2:15?”).
• No Audio Storage: The original audio file is discarded after processing, preserving only the sanitized text.

---

Error Handling & Troubleshooting

Scenario	Outcome / Action
File exceeds 25 MB	Upload rejected before processing. Reduce size and try again.
Unsupported format	Upload rejected. Convert file to a supported type.
Antivirus failure or detection	Upload blocked for safety. Provide a clean file.
Plan below Basic	Upload blocked with “not allowed” message. Upgrade plan to continue.
Transcription failure	Batch ends in error state. Retry later or contact support.

---

Best Practices

• Keep files short to stay under the 25 MB limit and reduce processing time.
• Use clear audio for more accurate transcripts (minimal background noise).
• Enable timestamps when you need to cite exact moments or create training data with segment metadata.
• Organize batches with descriptive titles to simplify retrieval and auditing.

---

Summary

Hana Audio Memory provides a secure, plan‑aware pipeline for turning spoken content into useful, searchable knowledge. By supporting common audio formats, optional timestamps, and antivirus scanning, it fits seamlessly into the broader memory ingestion system. Administrators can rely on it to capture important discussions and make them available across their organization, while maintaining governance and role-based access control.