Introduction

In an unconventional yet highly demanding transcription scenario, Jay records and transcribes dream reports as participants are just waking up. These recordings often contain hushed voices, background noise, and speech distortions. Conventional transcription tools proved inadequate, either lacking in accuracy, violating privacy, or requiring excessive manual cleanup. Jay sought a solution that was accurate, secure, and aligned with ethical AI practices.

1. Can you provide a brief overview of how you incorporate Alice in your workflow?

"I utilize Alice in a fairly unique and challenging setting when it comes to audio fidelity, I'm using it to transcribe audio recordings of users' dreams, made often as they are just waking up, without a great emphasis on clarity of speech, room tone, et cetera. Thus, confidentiality and as much accuracy as possible is critical."

2. What motivated you to start using Alice for transcription and which services had you tried before and why didn't those work?

"Alice's ethics statement and support of journalists were highly motivating. As I had used Otter in the past. Not only was it often inaccurate, its extremely dodgy privacy practices and AI training concerns left me with a very poor taste. I remain skeptical of AI in general, and use only in limited/highly specific scenarios. I also used Whisper as a plugin in Audacity, but its very slow processing and additional work needed to extract a usable transcript makes it unusable. The native iOS transcription service is very inaccurate."

3. How did these challenges impact your efficiency and workflow?

"As one can imagine, the poor accuracy is a significant barrier to use, especially when considering cost and any sacrifice made with regard to the ethics of the platform. Because of the challenging recording environment (often hushed or garbled voice, white noise, natural vocal chords distortion due to post-snoring, dehydration, etc.) recordings are often reviewed word-for-word for accuracy. Higher fidelity transcription reduces the time needed to spend on problem areas. Lesser products produce worse results, with greater losses in integrity in the overall process."

4. How was the process of integrating Alice into your existing systems and workflow?

"Very smooth, especially when we worked out kinks with the Dropbox/Google integrations. There was a period when folders were being made with null data, but Alice's engineers worked efficiently on a fix. The formatting of the rendered text is superior to other products, making this far less time consuming to integrate into our logs and format into intelligible and useful data."

5. How has Alice improved the processing speed and accuracy of your work? Anything that stood out as being especially important/noteworthy?

"By dint of comparison, I'll say that Alice is vastly superior and far faster than other methods which require ridiculous amounts of time correcting. Apple's native iOS transcription for example is "faster" only in that it is readily available, but so very often wrong. When we're working with hundreds of dream reports over months transcribed by iOS, we may acquire the pages of data faster but the accuracy is trash. Alice compares in accuracy only with Whisper on the Large V-2 model, which would take hours on a 15 minute recording."

6. Can you provide specific metrics or examples that highlight the improvements brought by Alice?

"This may not be the feedback you're looking for, but in the past several months I've had to make budgetary decisions due to a climate change related relocation and resulting changes in funding streams. This means I've not really been able to utilize Alice fully as I had, leading to a massively frustrating loss in time, energy and even files to faulty iOS system upgrades. I have been missing Alice's ease of use, and so have my participants. I have promoted Alice as a case study in workshops (your website throws a 404) as a tool for dream recall transcription as well as advanced it as a rare ethical use of AI in a professional clinical dream symposium, so I remain an ardent cheerleader. I believe Alice represents a rarity in the field especially when AI is so problematic."

7. Can you share any feedback or testimonials from your team about the experience with Alice?

"My only constructive feedback is that the cost structure is understandable but still a tad prohibitive. I know it is comparable/competitive to the others, and I'll not ever go back to Otter or another paid service under any circumstances. I cannot offer client testimonials without breach of confidentiality in this email, but I can reach out and ask for their direct testimonials and will anonymize. I'm not sure what screenshots to offer but happy to send any supporting ones along you request if you'd like to drill down on any of these. I only have 8 minutes of recording time available but I could easily demo a comparison between iOS native transcription/Alice regarding my use scenario."

Conclusion

Jay’s use case demonstrates Alice’s capacity to excel in one of the most demanding transcription environments — low-fidelity audio, high confidentiality requirements, and ethical concerns around AI. By delivering superior accuracy, speed, and data formatting, Alice not only improved workflow efficiency but also aligned with the user’s ethical and privacy standards. While cost remains a consideration, Alice’s performance has inspired Jay to advocate for its use in both workshops and professional symposiums, highlighting it as a rare example of AI applied responsibly.