Translate Audio voiceover failed or is stuck
What each voiceover error in Translate Audio means and how to fix it: short recording, long text, low balance, cloning failure or a job that seems stuck.
Updated November 2, 20264 min read
The voiceover is created when you tap play on the Translation tab. If instead of the audio you see an error or the progress doesn't move, find the message below. Errors in recognition and translation, before the voiceover, are covered at the end.
Error messages and what to do
The table lists the texts exactly as the app shows them, starting with the most common.
| Message | What it means | What to do |
|---|---|---|
| «Failed to generate audio» | the voiceover service didn't return audio | tap play again; if it repeats, choose ElevenLabs on the Voiceover tab to check |
| a top-up screen opens instead of the audio | the balance doesn't cover the voiceover; MiniMax also needs 180 tokens for cloning | top up on the screen that opens, or choose a cheaper voice; see Not enough tokens |
| «Recording is too short. Minimum duration for voice cloning is 10 seconds.» | MiniMax can't clone a voice from less than 10 seconds | use Qwen 3 (from 3 seconds) or a longer recording |
| «Audio is too short. Please provide at least 3 seconds of speech for cloning.» | Qwen 3 needs at least 3 seconds | record longer or choose a standard voice |
| «Could not detect clear speech in this recording. Please re-record with louder, clearer speech and try again.» | MiniMax didn't hear speech it can clone | record again closer to the microphone, without music and noise |
| «The translation is too long to voice with this option (… characters, limit is 10000). Try a shorter recording or a different voice.» | MiniMax voices up to 10,000 characters at a time | choose Qwen 3, ElevenLabs or Dubbing v2, or split the recording |
| «The original recording is not ready yet. Please wait a moment and try generating the voice again.» | the original file hasn't finished saving on the server | wait a few seconds and tap play again |
| «Voice generation took too long. Please try again.» | the voice service didn't answer in time | tap play again |
| «Voice cloning is unavailable right now. A standard voice was used.» | Qwen 3 cloning failed, the translation was read by a standard voice | this is a result, not an error; for a clone, try later or choose MiniMax |
| «Audio too short for voice cloning, standard voice was used» | the recording is too short for any clone | record at least 3 seconds for Qwen 3 or 10 for MiniMax |
If the voiceover seems stuck
Cloned voices take longer than standard ones, and the wait depends on the option. While it works, the player shows Cloning… or Dubbing….
| Option | Usual time shown in settings |
|---|---|
| ElevenLabs, OpenAI | ≈ 2 s and ≈ 5 s |
| MiniMax | ≈ 10 s |
| Qwen 3 | ≈ 20 s, longer for a long translation |
| ElevenLabs Dubbing v2 | about as long as the recording itself |
Cloning and dubbing continue on the server even if you leave the screen. Open the result again from History (clock icon in the top right corner), and the app picks up the status.
Translate Audio doesn't send a push notification when the voiceover is ready. Check the result in History.
If nothing changes for much longer than the times above, switch to ElevenLabs on the Voiceover tab and tap play: it shows whether the problem is in the chosen option or in the connection.
If the player reset after you edited the translation
This is expected. After an edit the old audio no longer matches the text, and the app shows «Text was edited. Regenerate audio to apply changes.» Tap play to voice the new text. With ElevenLabs Dubbing v2 edits are not used at all: «Dubbing uses the original recording, so edited text is not applied.»
Will I lose tokens for a failed voiceover?
MiniMax charges 180 tokens for cloning plus the text before it starts. If the MiniMax voiceover fails, these tokens are returned automatically. When you voice the same result again with MiniMax, for example with another speed, the 180 tokens are not charged a second time. Dubbing v2 is charged after the dub is ready. Check the charges in Settings → Usage History.
If recognition or translation failed
These errors appear before the voiceover, right after you choose a file.
«Processing failed. Please try again.»: tap Try again on the card. If it repeats, check the connection and choose the file again.
«Recording is too short. Please record at least 1 second.»: the recording is shorter than a second.
«File too large» with «Please choose a file no larger than 100 MB.»: upload the file by link with Paste link, up to 2 GB, or shorten it. Details are in File format not supported or file too large.
When to contact support
Write to support if the same error repeats with ElevenLabs, or tokens were charged and there is no audio. Tap Report a problem in the Feedback block under the result: the form attaches the job and the details of your device. If the voiceover works but sounds wrong, see Why the translated voice sounds robotic.
Contact support
Describe the problem and we will reply by email. Add the file name or job ID if you have one.