AutoFix

🤯 Hallucinations are well documented in ML models

• This improves the TTS quality by detecting TTS hallucinations and repeating the TTS request.
• This enables better quality audio all the time.

💼 How does it work 💼

• TTS assets are scored on an internal metric based on speech quality and background noise
• A threshold to detect audio hallucination has been set using internal data
• When an hallucination is detected, the TTS call is automatically repeated in ms-tts

Video Player

• We added a video player. So you can now share your audio with video 📹 💯

• We improved our asset cards and added some great bug fixes as well, to improve performance. 🐛

New voices 😀

We added a new vendor Cartesia they are working on new model types for foundational models 🤯

• Have a listen to their voices in the voice library under provider=cartesia
• You may want to try out chanel in 🇫🇷 , harshita for a great 🇮🇳 accent in English, and kane in 🇺🇸
• If you have a favourite please email us at support[at]audiostack[dot]ai we love to hear your feedback

New Email Design 📧

We gave a lick of paint to our emails - to delight our customers

We're excited to announce the addition of 34 new English-speaking voices to the AudioStack Library, upping the total number of available voices to a whopping 1407 speakers! 🎊

We made some updates to Workshop to better support different lengths of audio asset.

• The generate button will only be available if you have a script that fits (to avoid cutting off the end of your script); and,
• Editing a SonicSell Ad within Workshop now generates audio with the same length (bug fix).

Earlier this week, we released a couple of updates to the AudioStack platform, including:

• Redesign of the Voice Library - it's now easier than ever to find a voice in the AudioStack library.
• Mobile support for the Voice Library - browse voices from your smartphone. Simply open the page in your browser.

• Custom Length in Workshop - you can now specify how long you want an asset created in workshop to be.

• Fixed a bug with sorting in the Voice Library.

This week, the AudioStack team have been busy levelling up the capabilities of AudioStack API. We're excited to announce a new feature in Production - atmospheric sound design.

You can check out an example of what's possible below, the full tutorials will be landing shortly 🚀

We also did some maintenance to improve performance of the speech/voice endpoint and speech/voice/recommendation endpoint, as well as fixing a bug with load balancing in response to customer feedback.

This week in SonicSell, we added a couple of improvements to the ad generation experience.

• You can now "Start Over" to create a new ad; and,
• Voice selection is now more variable when gender and tone are pre-selected, improving the quality of voice recommendations.

We also improved the Voice Library by giving each voice its own page, so now you can link people directly to a specific voice in the library using the URL.

Some of our documentation has also had an update...

• We added more explanations and a better structure to “how does audiostack work”
• Added an errors page aimed at API users
• Enhanced the Getting Started page as well including adding a “what is a RESTful API” explanation.

SonicSell improvements

• Added some new example templates to the model to give the script generation even more expert guidance
• Length adjustments -> percentage of too short ads decreased from 50% to 15% 🚀
• Prompt refinements that should lead to less normalization errors

Infrastructure improvements

• Upgraded our user permission management system - should lead to less errors for our customers 😀
• Fixed a billing error on Speech to Speech 💰

JS SDK 1.1.13:

• Handle 200 & 202 responses correctly for STS (speech to speech)

This week, we added Speech-to-Speech to the AudioStack Platform, so you can now use this functionality without needing to write any code. Simply upload a speech file and select a voice, and you'll be able to generate synthetic speech that mimics your recording. Find out more in our guide: https://docs.audiostack.ai/docs/speech-to-speech

SonicSell

We also added a few improvements to SonicSell, including:

• Improved script editor performance; and,
• Fixed a bug encoding special characters.

As always, if you have any questions or feedback, please get in touch at support [at] audiostack [dot] ai.

Today we added a guide to make it easier to understand how organisations and users work at AudioStack, to help with getting the right role based access to workflows and resources.

Read more here.

We're also pleased to announce we added 15 new Japanese voices to the library in response to users' feedback.

Exciting news! You might've noticed that we recently took SonicSell v1.6.0 out of beta and made it available to all AudioStack users. SonicSell has undergone some major UX improvements as well as some exciting feature additions.

What's new?

• Improved ad quality: We’ve improved the length of ads to fit better, and the content of ads themselves.
• Improved UX: Based on feedback from users, we’ve made SonicSell easier to use than ever before. Create an ad in a single click, customise your ad by shaping your recommendations with a mood and tone of voice, or choose a voice and sound design for yourself.
• Give your ad a cover to make it stand out.
• All ads are now saved to your organisation’s library, so you can easily return to them later, and to enable collaborative editing. This also makes it easier to find your ads - search by name or content.
• Edit your ad in the workshop: Tailor your ad to fit a specific audience or a brand tone of voice, or enhance your ad. You can also create an ad from scratch in the Workshop using your own script.
• Create an ad with no sound design: This is perfect for situations where you want to download just the voice, to combine with other audio in a DAW or video editing software.

There are even more features still to come - watch this space for updates in the coming weeks. For the next few weeks, the old version of SonicSell will stay available alongside the new version. We will deprecate this at some point in the near future, so please try out the new version and let us know ASAP if there is anything you could do on the old version but aren’t able to do here.

You can find our complete user guide here: https://docs.audiostack.ai/docs/sonicsell-v160