Ask HN: How to boost Gemini transcription accuracy for company names?

#1

Post by bingwu1995 » Mon, Oct 20, 2025, 4:41 AM UTC

I’m using Gemini for speech-to-text and it often misrecognizes company names and acronyms.

Is there any way to use a custom lexicon or vocabulary with Gemini to improve recognition accuracy? If not directly supported, what are practical workarounds people use — e.g. preprocessing prompts, fine-tuning, or combining Gemini with another ASR that supports phrase boosting?

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#2

Post by gallexme » Sun, Oct 26, 2025, 4:35 PM UTC

Adding it to the instructions worked well for me with specific terms

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#3

Post by semessier » Sun, Oct 26, 2025, 4:35 PM UTC

adding to the question, ruling out fine-tuning for practicality, what about injecting names towards the embedding but not into the context?

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#4

Post by lysecret » Sun, Oct 26, 2025, 4:50 PM UTC

I generally found 4o-transcribe to be more performant than gemini fyi.

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#5

Post by Reubend » Sun, Oct 26, 2025, 5:07 PM UTC

Any company names or special acronyms should be added to your prompt.

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#6

Post by tifa2up » Sun, Oct 26, 2025, 5:41 PM UTC

Don't solve it on the STT level. Get the raw transcription from Gemini then pass the output to an LLM to fix company names and other modifications.

Happy to share more details if helpful.

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#7

Post by huflungdung » Sun, Oct 26, 2025, 5:44 PM UTC

[dead]

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#8

Post by simonw » Sun, Oct 26, 2025, 5:46 PM UTC

Have you tried feeding it a system prompt with a list of custom vocabulary? I would expect that to work really well.

"Transcribe this audio. Be careful to spell the following names and acronyms right: list-goes-here"

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#9

Post by remus » Sun, Oct 26, 2025, 5:48 PM UTC

tifa2up wrote: post #6

Don't solve it on the STT level. Get the raw transcription from Gemini then pass the output to an LLM to fix company names and other modifications. Happy to share more details if helpful.

I've had some luck with this in other contexts. Get the initial transcript from STT (e.g. whisper), then feed that in to gemini with a prompt giving it as much extra context as possible. For example "This is a transcript from a youtube video. It's a conversation between x people, where they talk about y and z. Please clean up the transcript, paying particular attention to company names and acronyms."

Re: Ask HN: How to boost Gemini transcription accuracy for company names?

#10

Post by samtts » Sun, Oct 26, 2025, 5:58 PM UTC

[dead]