Troubleshooting
Find your symptom below: the upload will not finish, the session failed with an error, or the notes look wrong. Most issues come down to the audio file itself.
My Upload Won't Finish
-
How to fix it: Large files take a few minutes on slower connections. Check your internet connection and try again. If it keeps failing, try a different browser (Chrome or Firefox work well) or clear your browser cache. For a very large file on a slow connection, converting it to a smaller MP3 with a free tool like Audacity makes the upload much faster.
-
What it means: Your upload is over the 4GB limit. This is checked in your browser before the upload starts, so you'll see a message like "File '...' is too large (X MB). Individual files must be under 4GB." for a single oversized file, or "Total size (X MB) exceeds 4GB limit." for the combined upload.
How to fix it: Export to a compressed format like MP3 or OPUS to shrink the file, or split a long recording into multiple sessions. Uncompressed WAV files are the most likely to exceed the limit.
My Session Failed with an Error
-
What to expect: A 2 to 3 hour session usually finishes in 5 to 10 minutes. Longer sessions or heavy load can stretch that to 20 to 30 minutes.
How to fix it: Refresh the page first, since the status indicator sometimes needs a manual refresh to catch up. If it is still processing after 30 minutes, click the Feedback button inside the app.
-
Some failures are temporary and are not always shown with a detailed message. Support will be notified and will investigate the issue.
How to fix it: A retry is usually all it takes, and if it keeps failing, report it with your session ID. If you were charged for an analysis that never completed, contact us through the Feedback button or at support@gmassistant.app and we will put the credits back.
-
What it means: The file you uploaded is not one of the audio formats GM Assistant can read.
How to fix it: Re-export your recording as MP3, M4A, WAV, FLAC, OGG, or OPUS. MP3 is the safest choice and processes fastest. A free tool like Audacity can convert most files.
-
What it means: The file is a supported format on paper, but the audio data could not be read. This usually means the recording was truncated or the file was damaged during transfer.
How to fix it: Play the file locally to confirm it works. If it does, re-export a fresh copy and upload again. If the original recording itself is cut off, you may only be able to recover the portion that plays.
-
What it means: Recordings must be at least 1 minute and no more than 10 hours long. Files outside that range are rejected before analysis. The exact message is either "Audio duration of X seconds is too short. 1 minute is the minimum duration." or "Audio duration of X hours is too long."
How to fix it: For a very short clip, combine it with the rest of the session. For a recording longer than 10 hours, split it into parts and upload each as a separate session. See Uploading Audio.
-
What it means: The recording was scanned for speech and none was found. The whole file looked like silence, background noise, or music with no audible talking.
How to fix it:
- Confirm you uploaded the right file, and that it is the track with people talking rather than a music-only or muted channel.
- Play the file and check the volume. Very quiet recordings can read as silence.
- If you recorded multiple tracks, make sure the speech track is the one being uploaded. See Uploading Audio.
-
What it means: Speech was present but nothing recognizable could be transcribed from it. This usually points to very quiet, muffled, or heavily distorted audio, or the wrong language setting.
How to fix it: Review the Recording Audio guide for better capture, confirm the recording language matches what was spoken, and try again with a cleaner file.
-
What it means: The transcribed speech rate came out implausibly fast for natural conversation.
How to fix it:
- Sped-up audio: Upload your recording at its original 1x speed. Do not run it through a speed-up or time-stretch tool, and avoid podcast apps that export at increased playback speed. Re-export the original recording at normal speed and upload that.
- Multiple parts uploaded as separate tracks: If your session was recorded in pieces, those pieces are parts of one recording, not separate tracks. Uploading them as multi-track lays them over the same timeline and transcribes them on top of each other. Upload them as multi-part audio instead, using Add Part so they play in sequence.
- The same voices in multiple tracks: Multi-track expects each track to hold a different speaker. If your tracks each captured everyone (for example several mics in the same room), every voice is transcribed once per track and the word count multiplies. Upload a single mixed-down file, or pick one track, rather than uploading the overlapping tracks as multi-track. See Multi-Track and Speaker Labels for when multi-track is the right choice.
-
What it means: You don't have enough credits to analyze this session at the level you chose. The message shows how many credits the analysis needs against your current balance.
How to fix it: Buy more credits from Purchase Credits in the menu, or switch from Comprehensive to Brief, which costs fewer credits per hour. You will need to create a new session to retry. See how charging works for the per-hour rates.
-
What it means: The content safety filter on the analysis model blocked the response. The full message is: "We couldn't generate an analysis for this session, possibly due to content that was flagged as inappropriate. If that seems to be in error, please contact support."
The filters are deliberately loosen for tabletop play, so in-character villainy, threats, hate from a villain, and combat descriptions normally pass fine. A session is only blocked when it contains:
- Sexually explicit content at a high level of severity.
- Dangerous content at a high level of severity, such as detailed real-world instructions for serious harm.
- Anything involving the sexual abuse of minors.
How to fix it: If you believe the block is a mistake, use the in-app Feedback button or email support@gmassistant.app with your session ID so we can look into it.
My Notes Look Wrong
-
How to fix it:
- To fix the session you already have, use Smart Edit to correct misspelled names and other issues in the analysis.
- For future sessions, add your characters, locations, and items in Campaign Details so spelling correction can fix proper nouns automatically.
- Poor audio is the usual cause of garbled text. The Recording Audio guide covers simple fixes, like where to put the microphone.
- When players excessively talk over each other, the transcript blends voices together. Encourage one speaker at a time, and use headphones for online games.
-
- Important events were missed: If you used Brief, re-analyze at Comprehensive for more detail.
- Wrong characters or locations: Edit the fields directly in the Notes tab, and use Find & Replace to fix a recurring misspelling across the whole session.
- Output in the wrong language: Confirm the recording language is set correctly, and contact support if the analysis comes back in a different language than was spoken.