Summary or Recap Not Generating

If a Hedy session’s summary or recap won’t generate — it shows an error, gets stuck loading, or just never appears — the first thing to do is tap the Re-generate Summary or Update Summary action on the session screen. Most summary issues are resolved by a single retry. If retrying doesn’t work, the cause is usually local AI hitting a context limit, a long transcript timing out, or a network issue preventing cloud AI from reaching its provider.

First Step: Retry the Summary

On the session detail screen, look for the Re-generate Summary action (or Update Summary — the label depends on context). This sends the existing transcript back to your configured AI provider and produces a fresh summary without re-recording anything.

A retry resolves most cases because failures are often transient — a momentary network hiccup, a busy cloud provider, or a one-off processing glitch.

Cloud AI vs. Local AI: Which Is Generating Your Summary?

Hedy can generate summaries either in the cloud or locally on your device. This is controlled by a setting:

Cloud AI (default for most users) — sends the transcript to a remote AI provider. Needs an internet connection. If your primary cloud provider hits an outage, Hedy may fail over between Portkey and Vertex on iPhone, iPad, Android, and Mac. On Windows, Portkey has no Vertex backup, so an outage there means the summary fails.
Local AI — runs the AI model directly on your device. Works offline. Requires you to have downloaded a local AI model. There is no cloud fallback when local AI is enabled — if the local model fails, the summary fails.

You can check which is active in Settings > Speech & AI > Local AI Processing. If you’ve enabled Local AI but you’re trying to summarize a very long session, you may be hitting the local model’s context limit.

”This Request Is Too Long for the Local AI Model”

If you see this message, your transcript is too long for the local AI model you’re using. You have two options:

Switch to a larger local model. Bigger models can handle longer transcripts but require more memory and disk space. See Local AI Processing.
Switch to cloud AI for this session. Open Settings > Speech & AI > Local AI Processing and turn it off. Then re-generate the summary. Cloud AI providers handle much longer transcripts than typical local models.

This message only appears with local AI — cloud AI can usually handle 4+ hour transcripts without hitting context limits.

”Failed to Generate Summary. Please Try Again.”

This generic error means the summary attempt completed without producing usable output. Common causes:

Network interruption during the request — retry once your connection is stable
AI provider rate limit or outage — retry in a few minutes
Transcript content that the AI provider’s content moderation flagged — rare, but happens with certain medical or legal content
Timeout — summary generation has a 15-minute maximum runtime. Sessions over ~4-6 hours can occasionally exceed this on a slower local model.

The first thing to try is always Re-generate Summary. If two retries in a row fail, switch to the other AI mode (local ↔ cloud) and try once more.

”Processing Didn’t Complete”

If a session shows “Processing didn’t complete” with a Complete processing action, the post-session AI work was interrupted before finishing — most commonly because the app was closed too soon after a session ended, or the device lost power mid-processing. Tap Complete processing to resume.

Verify the Transcript Before Blaming the Summary

If a summary looks wrong (not failed, just wrong — missed topics, wrong people, off-key content), check the underlying transcript first. Open the session and scroll through. If the transcript itself is missing chunks, has wrong words, or is in the wrong language, the summary will inherit those problems regardless of how good the AI is.

For missing transcript chunks, see Transcription Stopped Mid-Session
For accuracy issues, see Fix Poor Transcription Quality
For wrong-language transcripts, see Transcript Came Out in the Wrong Language

Fix those first — re-generating the summary against a correct transcript usually solves the rest.

Summary Generation Takes Too Long

Summary generation has a 15-minute maximum. If you’ve waited longer than that and nothing has appeared:

Tap Re-generate Summary — the previous attempt may have timed out silently
If you’re on local AI for a long session, consider switching to cloud AI temporarily
If you’re on cloud AI, check your network — slow uploads can prevent the request from reaching the AI provider in time

For typical 30-60 minute sessions, summaries should complete within 30-90 seconds.

What’s in a Summary

When working correctly, Hedy’s recap contains: a session title, a high-level summary, key points, action items, and any to-dos addressed to you. If any specific section is missing but others are present, that’s usually a sign the AI model decided that section wasn’t applicable — not a generation failure.

Detailed topic-by-topic notes are generated by a separate service, not the recap pipeline. If notes are missing but the summary is fine, that’s a different issue.

Editing Transcripts and Regenerating Content
Local AI Processing
Hedy Can’t Connect (VPN, Firewall, Corporate Network) — if cloud AI is unreachable
Fix Poor Transcription Quality — fix the transcript before regenerating the summary

Still having trouble? Contact us with: which AI mode you’re on (cloud or local), the exact error message, the session length, and whether retrying resolved it.