Since the last update in 2.5 flash tts i’m getting some issues with multispeaker generations. Sometimes the voice is the same for both speakers for the whole audio or just for a part of it.
Hi @Herian ,
Welcome to the Forum!
Thank you for reporting this issue.
This seems to be a temporary issue, could you please try again and let us know if still the issue perists?
The issue is still there, not for every generation, but it happens a lot. In addition, the longest is the output, the more is degraded. After something like 2mins, there are crackling sounds, noises, the model ignores the directors notes (or say it instead) and the voice swap issue is more noticeble.
@Mrinal_Ghosh any news on this issue? There’s some update planned? It’s really unreliable by now
I too am facing this problem. Sometimes, both the speakers ending up with same voice - it’s infrequent, but it is real. Faced this across days in last few days, so it’s not a one-off that will work with retry or refresh.
Another problem - occasionally, if there are 2 female voices, the second speaker talks in male voice.
Is there an earlier version of the model (pre-December changes) that we can invoke? Until you solve the problem, we can continue with the earlier version because this is impacting us a lot.
I haven’t faced any issue anymore, but i haven’t used it a lot…
Hi @Herian,
Thanks for following up! We have escalated this issue to the concerned team for further investigation.