ProZ.com Translation Forums

Support for local LLMs in Hera AI

InText — Wed, 24 Jun 2026 07:56:00 +0000

Forum: Hera AI
Topic: Support for local LLMs in Hera AI
Poster: InText

No more data privacy trade-offs — Hera AI now supports local LLMs!

If you’re working with sensitive content (legal, financial, medical translations), you can connect Hera AI to a local LLM (for example, via Ollama) so that all files stay within your own infrastructure.

Working with a local LLM means that:
🔹File processing happens on your own server (not a third party’s)
🔹Your data isn’t used to train third-party models
🔹You stay in full control of which LLM you use and how it’s configured

Hera AI also supports hybrid workflows: use local models for sensitive projects, and use OpenAI cloud models for everything else.

Want to test this setup on your projects? Try Hera AI for free: [url removed]

Hera AI: What it is, how to install it, and how to use it | Local models support

InText — Tue, 23 Jun 2026 10:56:00 +0000

Forum: Hera AI
Topic: Hera AI: What it is, how to install it, and how to use it
Poster: InText
Post title: Local models support

[quote]Robert Pätzold wrote:

Привіт,

this sounds really interesting.

How about support for Gemini and für local models (Qwen 3.6, Gemma 4)?

Best regards
Robert [/quote]

Hello Robert,

We've released a new version of Hera AI that includes one of the features you asked about earlier — support for local LLMs via Ollama, which allows you to connect models like Qwen or Gemma directly to Hera AI.

If that's still relevant for you, let me know and I'll send over a trial license.

As for Gemini support, it's still not in our immediate pipeline, but it remains on our roadmap.

Best regards,
Maksym

Hera AI: What it is, how to install it, and how to use it | Gemini and local models support

InText — Wed, 22 Apr 2026 07:39:00 +0000

Forum: Hera AI
Topic: Hera AI: What it is, how to install it, and how to use it
Poster: InText
Post title: Gemini and local models support

[quote]Robert Pätzold wrote:

Привіт,

this sounds really interesting.

How about support for Gemini and für local models (Qwen 3.6, Gemma 4)?

Best regards
Robert [/quote]

Hallo Robert,

Thanks for reaching out! Great timing — we're releasing Hera AI 3.0 next week, which will include support for local LLMs via Ollama. That means you'll be able to connect models like Qwen or Gemma directly to Hera AI.

As for Gemini, it's on our roadmap, though not in the immediate pipeline.

Once Hera AI 3.0 is out next week, you're welcome to sign up for a free trial on our website: [url removed]

Best regards,
Maksym

Hera AI: What it is, how to install it, and how to use it | Support for Gemini and für local models (Qwen 3.6, Gemma 4)

Robert Pätzold — Tue, 21 Apr 2026 21:58:00 +0000

Forum: Hera AI
Topic: Hera AI: What it is, how to install it, and how to use it
Poster: Robert Pätzold
Post title: Support for Gemini and für local models (Qwen 3.6, Gemma 4)

Привіт,

this sounds really interesting.

How about support for Gemini and für local models (Qwen 3.6, Gemma 4)?

Best regards
Robert

Hera AI: What it is, how to install it, and how to use it

InText — Thu, 26 Mar 2026 16:29:00 +0000

Forum: Hera AI
Topic: Hera AI: What it is, how to install it, and how to use it
Poster: InText

Hi everyone :)

My name is Maksym, I’m the head of localization at InText. This is the official support forum for [b]Hera AI[/b] — a tool built to support linguists working with machine-translated content.
A bit of context: at InText, we have been working with freelance linguists for decades. Hera AI came out of that experience — especially from the day-to-day challenges of MTPE workflows.

[b]What Hera AI is[/b]
If you work with MTPE, you know the challenge: MT output that ignores your glossary, misses the client's tone, and has to be corrected segment by segment without sufficient context.
Hera AI is designed to improve that process.
It works directly with bilingual file formats, applies your reference materials (style guides, glossaries, TMs, etc.), and applies context across multiple segments at once, rather than processing them one at a time.

[b]Key features:[/b]
[list]
[*] [b]Automated post-editing[/b]. Automatically post-edits machine-translated bilingual files based on reference materials.
[*] [b]Automated linguistic quality assurance (LQA)[/b]. Reviews bilingual files against customizable LQA rules defined for a particular project or language pair, ensuring compliance with your quality standards. Operates in comment mode, leaving the original translations unchanged.
[*] [b]Automated quality estimation (QE)[/b]. Evaluates bilingual files according to TAUS Post-Editing Guidelines and identifies which segments require light, full or no post-editing. Operates in comment mode, leaving the original translations unchanged.
[*] [b]LLM-based translation[/b]. Generates translations directly from the source file into the target language while maintaining quality through contextual understanding across multiple segments.
[/list]

[b]How to install Hera AI[/b]
[List=1]
[*] Visit the [url= [url removed] ]Hera AI page[/url] and fill out either the “Try a 30-day free trial” form, or the “Purchase Hera AI” form to access the free Basic plan. The Basic plan supports up to 3 target languages and is well suited for freelancers.
[*] You will receive an email with a download link and a license file within 1 hour. This is a manual step, so please allow a little time.
[*] Download the .zip archive from the link provided and paste the license file into the Hera AI folder.
[*] Open the .exe file and you are ready to begin.
[/list]

[b]How to use Hera AI[/b]
The standard workflow is as follows:

[list=1]
[*] Upload your bilingual files.
Supported formats: .sdlxliff, .mqxliff, .mxliff, .xliff, .xlf (Trados, memoQ, Phrase, Crowdin, Lokalise, Smartcat, Smartling, Wordbee, and XTM), and .tmx.
[img] [url removed] [/img]
[*] Upload your reference materials and project-specific instructions to guide terminology and style.
[img] [url removed] [/img]
[*] Select an AI Assistant for your task — post-editing, LQA, quality estimation, or translation — or create a custom one (required if you want AI to apply your reference materials).
[img] [url removed] [/img]
[*] Choose how many segments should be processed together to improve contextual accuracy.
[img] [url removed] [/img]
[*] Receive improved output in the original file format, ready for further linguistic review.
[img] [url removed] [/img]
[/list]

For a step-by-step instruction, please refer to the [url= [url removed] #heading=h.cqwqihljb8gv]Hera AI Guide[/url].

[b]Privacy & security[/b]
I know this is a common concern, so I want to address it upfront.
Hera AI does not store your input data or use it for training purposes. Your uploaded files are transmitted directly from your device to OpenAI for processing. OpenAI guarantees that content submitted through its API is not stored, logged, or used to train models — all processing happens in a secure, one-time interaction, so your content remains fully confidential.
If needed, you can also disable data sharing in your OpenAI account settings.

I would be glad to welcome ProZ members among our users! Please use this forum to ask questions, report any issues, or share suggestions for improvement — all feedback is welcome.

Best regards,
Maksym Hamii
Head of Localization
InText

[Edited at 2026-03-26 16:36 GMT]

[Edited at 2026-03-26 16:38 GMT]