Deepgram Reviews & Product Details

I use Deepgram for medical transcription, and Nova 3 is particularly capable at recognizing medical pronunciations, including disease and medicine names that other models miss. I find the transcription model best in class, and its accuracy is better than the OpenAI speech model I used before. The interface is easy to use, so I can explore the models without much navigating. The service is reliable, with high uptime, and the cost is pretty nominal. Setup was pretty easy and took five minutes to get started. Review collected by and hosted on G2.com.

I have some trouble managing keys. Our system has security issues, so we have to generate a Deepgram API key for each person, and sometimes Deepgram fails to generate a new key for team members. That's the only issue we've faced. Review collected by and hosted on G2.com.

I liked that Deepgram was easy to use and fast, and that its transcriptions are almost 100% accurate. It saved me a lot of the process and time: it gave me a good base to start transcribing, instead of leaving me with a blank canvas. I also liked being able to divide the transcription by speakers, the confidentiality, that it is not expensive, and that it has many useful advanced features. The setup was quite easy for me, with the teacher's instructions, and I felt it lightened this type of work a lot. Review collected by and hosted on G2.com.

Although Deepgram allows for speaker differentiation, sometimes it doesn't do so accurately enough: it mixes or merges paragraphs from different people and can attribute them to a single speaker. Most of the time it works well, but that part could be improved. Review collected by and hosted on G2.com.

I like the voices, especially Ryan and Daniela in Spanish, and I like that they sound human rather than robotic. Deepgram is quite fast, and I like the balance between the latency and the voice. It converts text into voice quickly, which matters when I’m driving and can’t read. Setup was quite straightforward: I registered, set up one or two voices and an API key, and let AI handle the rest. It hasn’t been that expensive as far as I know, and I think it works quite well in English. Review collected by and hosted on G2.com.

I find the Spanish voice options quite limited. Since I'm from Argentina, I'd like a more Latin Spanish voice. Right now, I'm stuck with Spanish from Spain, which sounds weird to me because I'm not used to it. Review collected by and hosted on G2.com.

I like that Deepgram is affordable and offers multiple models. I can test and benchmark different models for our healthcare-related work and choose the right one. Its voice transcriptions are fast and inexpensive, and sometimes better quality. The initial setup was pretty easy. Review collected by and hosted on G2.com.

I find Deepgram’s AI agents are still pretty early and don’t perform as well as ElevenLabs in terms of realism. The voices can sound robotic, and the models behind the writing can produce generic, robotic sentences. I’d like better, more realistic voices for Deepgram’s AI agents. Review collected by and hosted on G2.com.

The latency is extremely low, and the accuracy of the interim transcripts is impressively fast. Review collected by and hosted on G2.com.

Models like flux have different endpoints and support different versions of WebSocket connections. I think every model should share a universal connection so integration is more consistent. Review collected by and hosted on G2.com.

I use Deepgram for both speech-to-text (STT) and text-to-speech (TTS) in AI applications and workflow automation. It provides accurate transcription across multiple languages and produces natural-sounding speech with low latency, making it suitable for real-time and batch processing. The biggest advantage has been its language support and output quality. The transcription accuracy is consistently high, including for accents and multilingual content, and the TTS voices sound clear and natural. The API is straightforward to integrate, the documentation is well organized, and the response times are fast. Overall, Deepgram has been a reliable solution for building voice-enabled applications and automating audio processing tasks. The initial setup is also very easy. Review collected by and hosted on G2.com.

Deepgram performs very well overall, though there are a few areas that could be improved. I would like to see broader language and dialect support, particularly for less commonly used languages and regional accents. The TTS voice selection could be expanded with more natural and expressive voices across additional languages. Pricing for high-volume usage could also be more competitive, especially for developers building large-scale AI applications. More built-in features for speaker diarization, custom pronunciation, and voice customisation would make the platform even more versatile. Review collected by and hosted on G2.com.

I use Deepgram for cheap transcription, and the accuracy is at a decent enough level most of the time. I like the easy-to-use portal, especially how I can track API key usage and spending. The playground tool is also helpful for processing one-off audio recordings, and I've used this 4 or 5 times. The initial setup was good, and I believe I could sign in with Google. Review collected by and hosted on G2.com.

I think they do a lot with grants for bigger businesses, but it would be good to see more funding for SME businesses. Like with our SaaS product, if we had been offered say £1000 in credit, we would probably have used that to perfect our product and then had more customers and volume to put through Deepgram after the trial. Also, accuracy could do with a tweak in some areas, as it gets a lot of standard English names wrong. Review collected by and hosted on G2.com.

I like that Deepgram is very reasonably priced because you pay as you go and it's not a subscription model. It also has nice guardrails to prevent over purchasing or getting overcharged. Deepgram is highly responsive and very accurate in transcription and detection. I really appreciate that the live transcript delay is minimal, allowing me to follow along conversations easily. Setting it up was a breeze; creating an API and plugin was very easy. It also works really well with the apps I use for AI to plug in for live transcript. Deepgram is far faster and more accurate compared to the native transcription tools I've used in other AI programs. The amount of transcription I can do for the price is very competitive. I was very impressed with how super easy it was to set up using the secure API; it took literally minutes. Review collected by and hosted on G2.com.

No issues from my perspective, considering how well Deepgram works and how portable it is. Review collected by and hosted on G2.com.

I find Deepgram to be incredibly valuable, especially due to the transcription speed combined with its accuracy. It performs well even with accents, background noise, or overlapping speakers, which is crucial for AI services like Wirevox AI that we run. The real-time latency is noticeably better than a lot of competitors. The voice quality offered by Deepgram is outstanding and not many services offer such high quality. We also use their speech-to-text and text-to-speech models alongside different LLMs, integrations, and tools, and it's been seamless. The initial setup was incredibly easy, and though help was offered by Deepgram's team, the detailed documentation meant we didn't actually need it. Our users love Deepgram, as it offers better accuracy, latency, and cost efficiency at scale compared to what we used before. Review collected by and hosted on G2.com.

We would really love to have more French accent/language voices. As of right now, the only available French voices are Agathe and Hector, which aren't bad at all. They are amazing, but our users would really love more options to choose from. Review collected by and hosted on G2.com.

I’ve used Deepgram to create first cuts of PRDs, and it has fast-tracked progress for the product team. Review collected by and hosted on G2.com.

If the platform’s UI could be made a bit easier to use, especially from a user’s perspective, that would be nice. Review collected by and hosted on G2.com.