How to download a model for offline use

[AHeinlein]

I am looking for a way to get SpeechNote running on a machine without internet connection. Specifically, I am having problems getting models to work this way. I downloaded a model an a machine with SpeechNote and internet connection and got a file "multilang_whisper_large3_turbo.ggml" in "~/.cache/net.mkiol/dsnote/speech-models". I copied it over to the other machine into a folder "~/speech-models/" and pointed the SpeechModel installation there to this folder for speech models. I can see using strace that SpeechNote is looking for models in this folder on startup, and it picks up this file and opens it. Still, SpeechNote says there are no available models and I need to download one.

So obviously I'm missing something, but I can't find out what.

[mkiol]

Hi. Thank for the question.

Indeed, if you only copy the model file, it will not be immediately visible as "enabled". To make it "enabled" you can use one of the following methods:

• Simply click "Download" on the model you have copied. If the file is correct (the checksum matches), Speech Note will recognize that the file has already been downloaded and will simply mark the model as enabled without downloading, or
• you can copy an entry enabled_models from the configuration file (~/.config/net.mkiol/dsnote/settings.conf:

Example:

...
[service]
...
enabled_models=en_whisper_futo_english244, en_fasterwhisper_crisperwhisper, en_kokoro_82m_am_adam, en_sam_en, en_rhvoice_slt, en_f5_v1_base, en_parler_mini_expresso, en_bergamot_enja, ja_bergamot_jaen, en_bergamot_enzh, zh_bergamot_zhen, ja_kokoro_82m_jf_nezumi, zh_kokoro_82m_zf_xiaoyi, en_coqui_xtts202, sv_piper_nst_medium, sv_bergamot_sven, en_bergamot_ensv, de_fasterwhisper_crisperwhisper, sv_fasterwhisper_kblab_tiny, sv_fasterwhisper_kblab_base, en_bergamot_enar, ar_bergamot_aren, en_bergamot_enko, ko_bergamot_koen, en_bergamot_enpl, pl_bergamot_plen, sv_coqui_fairseq_swe, sv_coqui_cv_vits, pl_vosk_small, en_whisper_futo_english74
...

[Ivan2getdmodelz]

Can you please specify where the model files are stored? I've recently installed this via flatpack, and I am also trying to port and enable the downloaded language models, or add more models offline, but I don't know where to place the on the offline machine. Thanks in advance for the help

[mkiol]

@Ivan2getdmodelz The default location for Flatpak is in ~/.var/app/net.mkiol.SpeechNote/cache/net.mkiol/dsnote/speech-models. You can change it in the Settings->General->Location of language file.

[AHeinlein]

Indeed, if you only copy the model file, it will not be immediately visible as "enabled". To make it "enabled" you can use one of the following methods:

* Simply click "Download" on the model you have copied. If the file is correct (the checksum matches), Speech Note will recognize that the file has already been downloaded and will simply mark the model as enabled without downloading, or

Thank you for your help. That actually worked, except for the crisperwhisper model, which insists on displaying an EULA which fails. I got it to work using the second method, so no more problems on my side. Just wanted to let you know.

[mkiol]

Right, I forgot the "license dialog". The content of the license is fetched from the internet, so you can't activate the model when you are offline. I will fix this in the next version. Thank you for spotting this.