5.0 - New login methods

The login process for the Generator has received an overhaul. If this is your first time accessing after July 2019, please use the "New access?" link. You will not lose any features, but the new process will enable you to manage your own account. And it simplifies the process of adding new users.

4.0 - Combined marker files

All timing markers for each item are now combined into a single JSON file. This will provide performance improvements in TN8 because only one file will need to be downloaded per item, instead of one XML per access element per item.

3.7 - SSML marker updates

Support for converting more SSML rate controls to iSpeech speeds.

3.6 - SSML marker updates

iSpeech marker timing lags when using SSML rate controls on text, such as:

<prosody rate="1">hello world</prosody>

3.5 - SSML for webapp

SSML functionality from the API version of the generator has been incorporated into the webapp (this) version.

Updated iSpeech errors

There is a new error message for instances when iSpeech complains about invalid characters in a TTS request.

Admin console added

An admin console has been added where specified users can monitor recent generation jobs, check for frequent errors, and add new user accounts.

Updated timing/logging info

Each step in the generation process is more carefully timed, and this timing info is logged in the database so that we can more quantitatively monitor slowness.

Increased capacity

The TTS Generator environment has been updated to accommodate increased simultaneous usage.

Multiple voices - update

There was an issue in which multiple voices would not be generated unless the "Force refresh" option was used. This issue has been resolved.

Multiple voices

By selecting the "multiple voices" option, a female and male voice version of audio file will be generated. This will enable students to switch between male and female voices on any included item.

PNP Options

Any access elements included in Text Only or Text and Graphics will be generated.

Line breaks

The generator now removes any line break characters from the APIP text that is used for TTS. This will not affect line breaks that are in the item body, only in the non-visible APIP information. These were sometimes causing issues in the generation process and in the previewer.

iSpeech timing improved

iSpeech timing issues appear to be resolved!

Filename update

A suffix has been appended to new audio files that indicates which service, speed, and language were used for generation.

iSpeech retry

If the iSpeech server fails to respond, Genni was getting stuck and complaining that ogg's were failing to save. The error message is more accurate now, and Genni has better logic for retrying to contact iSpeech.

gh long passage improvements

gh has been tested to work for items that include long passages that are only one access element.

gh volume improvements

MP3 volume level is improved when generated with the gh service.

gh added

Speech service from gh has been added to the generator. The speed can be selected as with iSpeech, but Spanish voice is not available yet.

Default to 0

The iSpeech timing algorithm seems to be off for all speech rates except 0. For now, we will default to 0.

Force refresh

By selecting this option, audio files will not be retrieved from the database. They will be completely re-generated. This will be useful for cases in which the timing was incorrect at one point.

Spanish voice

When iSpeech Spanish voice is selected, the generator uses the Spanish pronunciation voice. This was broken.

APIP Validation

Reordered relatedElementInfo>spoken nodes in the access elements in order to follow the IMS APIP QTI specs.

Errors from ABBI-APIP

The Generator attempts to catch cases in which the APIP info has been corrupted. It will now throw an error instead of silently not generating the audio.

Backup speech service

The option to use IVONA for speech generation has been re-enabled.

Cosmetic updates

A few cosmetic updates, including the ability to close the file drop area if it gets stuck.

Major update

TTS Generator v.2 has been deployed. This update includes major speed improvements, especially when a package is running again because there are a few minor text changes. On the upload screen: error messages are more concise, and a timer appears next to the progress indicator. For the TestNav team, logging has been improved so that errors/issues are easier to find and track.

Making the jump to light speed

The generator processes have been separated into individual jobs that can be run simultaneously. This greatly increases the speed of running a package through the generator.

Backup speech service

The generator can now use IVONA to generate audio instead of iSpeech.

Generation insurance

A couple updates have been added to prevent failure and to double check that files are being generated accurately.

Caching with multiple users

Fixed an issue in which the generator would sometimes get stuck if multiple users were accessing the caching features at the same time.

Retry on failure

When one audio file fails to generate, the generator will retry up to two more times. If successful, the rest of the package will continue through generation. If it still fails after these three attempts, the generator will error out.

Retry audio generation

When one audio file fails to be generated, the generator retries two more times. If it is successful, then the progress is not lost. But if at that point it still fails, then the process needs to be restart.

Cleanup manifest

The generator removes references to old audio from the manifest, if a package is rerun through for TTS.

Orphan access elements

The generator removes references to access elements that are included in the inclusion order but no longer exist.

Audio not playing

Fixed an issue in which audio may not play after a package has been rerun through the generator after adding access elements.

Audio for IE

The generator now produces .mp3 and .ogg because IE had trouble playing .m4a.

Speech speed

On the upload page, choose the speed with which you want to generate speech.

Assets folders

Audio and xml files produced by the generator will be stored in folders called "assets."

Voice speed

The voice that reads text has been slowed down by about 20%.

Inclusion Order

Audio files are only generated for access elements that are included in the inclusion order, rather than generating all possible audio files.

Error Messages

Updated error messages to be more descriptive about where the issue is located in your xml files.

Invalid Characters

Generator checks for non-text characters (such as non-breaking spaces) so that an error message can be shown.