AssemblyAI Reviews

My primary use case was establishing a highly reliable video-to-text-to-content pipeline. AssemblyAI acted as the essential bridge between unstructured video data and a structured generative model. During integration, I realized that the quality of the downstream AI-generated formats depended on the accuracy of the initial transcription. If the speech-to-text API missed technical terms, the generated study aids were flawed. Using AssemblyAI ensured the transcript was highly accurate, meaning that the final educational tools generated by our LLM were of professional academic quality. Additionally, handling the asynchronous polling on our back end proved to be highly stable and easy to maintain.

Once my video is uploaded and turned into an MP3, AssemblyAI takes this MP3 file and converts it into text through its speech-to-text capability. This text is then fed into the AI. When the teacher logs into their dashboard, they fill a form with the lesson objectives and upload the MP4 video. As soon as this upload reaches our Node.js back end, I extract the audio and send it to AssemblyAI, which indicates the processing to the users. AssemblyAI works through the technical jargon. Within a minute or two, the teacher receives a notification that the lesson is ready, so they did not have to write the transcript or timestamp their video because AssemblyAI handled all the heavy lifting. For the student workflow, students enrolled in that specific teacher course open the lesson, watch the video, and then want to test their knowledge. Under the video player, they see generated flashcards, quizzes, or other study tools. Our platform does not need to reprocess the video. We take the high-accuracy text transcript already provided by AssemblyAI and feed it into our LLM to instantly generate 10 flashcards based exactly on what the teacher said in the video.

The best features AssemblyAI offers based on my integration experience include, first, the high-accuracy core transcription. It has the ability to accurately transcribe complex technical terminology including programming concepts and framework names and handles varying audio quality, such as classroom recordings with background noise, which is exceptional. The built-in file uploading through the /v2/upload endpoint is a huge time-saver for developers. It allowed me to stream audio files directly to their API for temporary hosting, eliminating the need to configure and manage intermediate public cloud storage such as AWS S3 before triggering transcription. The third feature is the precise word-level timestamps. The API returns the exact start and end times of every single word in the transcript. This metadata is essential for building modern e-learning features, such as synchronizing video playback with the transcript text or generating automated closed captions.

The integration of AssemblyAI has had a highly positive impact on my platform in three key areas. The significant faculty time savings means that automating the transcription process saved our instructors hours of manual labor per video. Instead of typing transcripts or drafting summaries, they could rely on the automated system, freeing up their time to focus on course quality and student interaction. The strong competitive advantage enabled us to launch our core adaptive learning feature set, transforming static, passive video lessons into interactive study tools including flashcards, quizzes, worksheets, and quiz games automatically, which sets our platform apart from standard video-only offerings. The last area is the low operational and infrastructure costs. Because AssemblyAI is a cloud-based, pay-as-you-go service, we avoided the high upfront costs of purchasing and maintaining expensive GPU hardware, which allowed us to offer automated study aids across our entire course catalog while keeping our margins highly efficient.

While AssemblyAI performs exceptionally well, there are a few areas where the developer experience could be further improved. First, regarding native video file support, currently, developers must write custom back-end logic to extract the audio track from video files locally before uploading. If AssemblyAI supported direct native video uploads and handled the audio extraction internally on their servers, it would simplify our back-end architecture. Native real-time status updates could also be improved because while the API is highly stable, writing custom asynchronous polling loops to check transcription status adds boilerplate code. Lastly, the queue latency for micro-files could be optimized because we noticed some initial queue or warm-up latency when transcribing very short audio files under one minute.

I have been working in my current field as a full stack developer and freelancer for nearly one year after I graduated from computer engineering.

AssemblyAI proved to be exceptionally stable throughout our development and testing phases. It has a high API uptime; I experienced near-perfect uptime on the public API endpoints. It maintains consistent response times and predictable HTTP status codes, with stable queuing and polling. The asynchronous transcription queue worked exactly as documented, where status transitions from queued to processing to completed never hung or failed silently, which made our Node.js polling logic highly reliable. It has robust connection handling, and we did not experience any connection resets.

AssemblyAI's scalability is excellent and requires zero infrastructure management from the developer because it relies on serverless resource scaling since it is a cloud-native API. It handles the scaling of GPU and CPU resources entirely on their end, so we did not have to worry about provisioning or scaling hardware to handle spikes in concurrent users. Its robust queue management means the asynchronous architecture handles spikes in concurrent transcription, where multiple uploads are placed in a stable queue and processed sequentially without crashing. The API is built to handle enterprise-level volumes, which means we can scale from a small local test environment to thousands of active students without making any changes to our back-end code.

Our experience with AssemblyAI's customer support and developer relations has been highly positive. It has excellent documentation and SDKs, responsive developer channels, and clear API error messages. For example, the API returns detailed, self-explanatory error codes and messages when our requests fail.

Before selecting AssemblyAI, I evaluated several other speech-to-text options, including the OpenAI Whisper API, AWS Transcribe, Google Speech-to-Text cloud, and the self-hosted open-source Whisper.

We saw a clear and immediate return on investment, both in terms of operational cost reduction and time saving. There was a 98% cost reduction on transcription because traditional manual human transcription costs a lot, so by using AssemblyAI, the cost dropped significantly. The instructor time saved means that manual transcribing lectures would take a lot of time, so using AssemblyAI saved massive chunks of time. Instant content generation also saved time for the students, so they do not wait for days for a teacher to manually write summaries and flashcards. Our automated pipeline generated study aids within two minutes of a video finishing its processing, dramatically improving the user experience.

Our experience with AssemblyAI's licensing and pricing was highly favorable because it has zero upfront fees. There are no licensing fees, setup costs, or long-term contract requirements. The cost-effective pay-as-you-go model means billing is strictly calculated per minute of audio processed, and the low barrier to entry with initial free promotional credits allowed us to build, integrate, and test our entire audio processing pipeline thoroughly without an upfront financial commitment.

Before selecting AssemblyAI, I evaluated several other speech-to-text options, including the OpenAI Whisper API, AWS Transcribe, Google Speech-to-Text cloud, and the self-hosted open-source Whisper.

An additional feature that deserves mention is the Auto Punctuation and Smart Formatting. This was highly valuable for our downstream generative AI pipeline because the transcript returned by AssemblyAI was already formatted as a clean written article. Our LLM, which is Gemini, could parse it easily, resulting in much higher quality generated summaries, quizzes, and flashcards for our students.

For other development teams considering AssemblyAI, I would offer the following advice based on our implementation: first, leverage the direct upload endpoint. During your initial prototyping and development, utilize the /v2/upload endpoint because streaming local files directly to AssemblyAI saved the overhead of configuring cloud storage buckets. Second, use webhooks for production; while writing a simple polling loop is easy for local testing, transition to their webhook notifications for production to save significant CPU and network resources on your back end. Lastly, plan the local media pipeline; if you are transcribing video files, ensure you build a robust and well-logged local audio extraction pipeline using tools such as FFmpeg to strip the audio track first, as this optimizes file transfer size and reduces processing latency.

AssemblyAI is the most effective tool that a developer could use, and I would rate this product a 9 out of 10.

I use AssemblyAI for audio transcription in multiple different languages. It has the capability of translating and transcribing into multiple different languages of both India as well as in the world. It also has good diarization capabilities, which is why I use AssemblyAI.

I had a customer use case problem where I had to transcribe lots of customer support calls into transcriptions in Hindi and multiple different Indic languages, as well as in foreign languages. AssemblyAI was helpful for this purpose.

AssemblyAI has been integrated into multiple different clients' use cases, and it was one of the core features in the AWS pipeline audio analytics pipeline that we created. It has benefited us significantly in saving costs of transcription.

The best features AssemblyAI offers are its blazing fast transcribing skills and accurate results. It also has the capability of diarization, as well as transcribing in multiple different languages, both in foreign and Indic languages.

I particularly value the accurate transcription of the language that the user provides as input and getting the best output without any kind of noise or silence. Automatic silence removal and voice activity detection are the best features of AssemblyAI that I appreciate in my daily use.

The outputs are really accurate. AssemblyAI already cares for the overall grammar, syntax, and the different nuances of the particular speakers. I believe the accuracy part has improved significantly from the previous versions that were available and should continue to improve further to become the best product in the market.

There was a saving of about 40 to 50% in transcription of audio analytics calls because previously, it was all done by humans, which could take days of effort and cost. This has significantly reduced to a great amount.

We tested with Deepgram and AWS transcription service that is already available in the market, and then we switched over to AssemblyAI.

AssemblyAI should definitely cater to multiple different languages of the world as well as in India. There are multiple different Indic languages and dialects available, and AssemblyAI should cater to those. Additionally, there might be multiple speakers available in a room in a particular meeting, and for that, proper diarization is required for identifying the different speakers as well as their names. These are some of the features that require attention by AssemblyAI, and they can definitely improve on that.

The pricing should definitely be looked at and the features should be worked upon as suggested.

I have been using AssemblyAI for about two to three years.

AssemblyAI is definitely stable.

AssemblyAI has a very good scalable solution. It has definitely been integrated in such a way that it handles multiple audios at a time. Regarding the pricing, I believe it is already in a very good range.

Customer support is definitely great with AssemblyAI. If you have any issues or encounter any problems in setting up, you can definitely reach out to the customer support and you can immediately get a solution.

I was using the AWS transcription service. There were problems of identifying the different languages, the different Indic languages that we have. AssemblyAI came into the picture and it solved a great deal of the problem.

The setup was pretty much easy. You just go to the AWS Marketplace and get this particular service provisioned and directly you can start using it with an API endpoint and key. The setup is pretty much easy.

I would say it is a time-saved and money-saved metric that should be considered here. That is how AssemblyAI is ruling the market.

I would give AssemblyAI a rating of 10 out of 10. I would suggest others to go for AssemblyAI because it is the best in the market in terms of accuracy, outputs, and the different languages that it caters to and transcribes. It is a very good product overall.

AssemblyAI has data privacy and security enabled so that the conversations that take place and are used for transcription are not leaked out to the public or leaked out in the public domain. There should not be any sort of sensitivity, privacy, or personally identifiable information data that gets leaked out. These things should be enforced strictly, and I believe AssemblyAI does that already.

Our main use case for AssemblyAI is automatically transcribing clients' meeting recordings, podcasts, and video interviews, and we also use it to generate summaries and extract key topics from long recordings. It saves our editor's team an enormous amount of time.

In one of my recent projects, we were producing weekly podcasts containing 12 different clients, and we had a meeting with the clients where we had to transcribe company show notes and repurpose them into blog content. Manually transcribing that volume was impossible for our small company, so we integrated AssemblyAI's API into our workflow, and within a minute of a recording being uploaded, it was fully transcribed and speaker-labeled. What used to take three hours per episode was reduced to under five minutes.

For client meetings, when we have the client meeting, some of us find it very difficult to note down the specific points and sometimes miss them, but by using AssemblyAI for that interview call, we get it easily transcribed. We have the main focus, and we get to know all the transcribed main points, so we don't miss out on anything.

We use an API integration to build AssemblyAI into our internal content management system, so when a file is uploaded, it automatically triggers the AssemblyAI transcription pipeline, and returns the result directly into our platform within minutes.

The best features AssemblyAI offers are the speaker diarization, which identifies who is speaking, the automatic summarization and sentiment analysis, topic detection, and the extremely accurate speech-to-text, even with different accents and background noise.

Speaker detection is what makes the biggest difference in my day-to-day work, especially when meetings happen with many people, multiple people interviewing, and panel discussions. It automatically identifies who the client is and who the speaker is, and for client-facing transcript accuracy, knowing who said what is absolutely critical, and AssemblyAI handles this better than any other tool we tested.

AssemblyAI has positively impacted our organization by allowing us to scale from managing five client accounts to 12 without hiring additional staff. Our client capability doubled while our costs stayed controlled, and client satisfaction scores also improved because the turnaround time on a transcript dropped from two days to same-day delivery.

AssemblyAI could be improved because the accuracy drops noticeably with a heavy accent or a very fast speaker, and pricing can become expensive at a high volume, so better multi-support or more affordable enterprise pricing tiers would make it significantly more competitive.

AssemblyAI takes data security seriously, offering data deletion options and not using submission audio to train their models by default, which is critical for us handling confidential client content. However, clearer documentation around compliance certificates such as SOC 2 and GDPR would give enterprise clients more confidence.

AssemblyAI is expensive, but overall, it is a good product.

I have been using AssemblyAI for about six months since joining the company.

AssemblyAI is stable in my experience; however, when the user's voice is unclear, it sometimes lags there.

Overall, the accuracy of AssemblyAI's output is consistently above 95% for clear audio, and it is reliable enough for professional use without heavy manual correction. The reliability of the API uptime has been excellent in our experience.

AssemblyAI's scalability can handle more volume if our company grows.

I never had to contact customer support because we never found any complaints or any bugs that would require us to contact them.

This was my first time using a transcribing application, and AssemblyAI did a great job.

We save approximately 85% of the time on transcribing tasks, and in workforce terms, we estimate AssemblyAI replaced what would have been a full-time transcriber role, which would cost around 35,000 to 40,000 per year. The API subscription costs a fraction of that, making the ROI extremely clear.

We saved around 85% of our workforce's time, and the cost savings are around 35,000 to 45,000 per year, making the ROI extremely clear.

AssemblyAI is a very good application for meetings, client interviewing, and podcasts, so I think everyone should use it in their company. I rate AssemblyAI an 8 out of 10 because the accuracy drops with heavy accents and fast speakers, and the pricing is expensive, so I think 8 is an appropriate rating for this application.

I use AssemblyAI mostly for the speech-to-text tool for chatbots. For example, I created Retrieval-Augmented Generation chatbots in which I used the microphone to ask questions to the bot, and I used AssemblyAI to convert speech to text and then send the text question to the LLM. This is my main use of AssemblyAI. Currently, my company also uses AssemblyAI for the same purpose, converting audio to text.

In my company, we have a chatbot to access questions about trade fairs and we have data from 15 years ago. We maintain a history going back 15 years, so we created the chatbot for that and we use AssemblyAI to convert audio to text. This allows users to communicate with the chatbot via audio instead of only in writing.

I have also used AssemblyAI for streaming, which is very useful. Currently, I am mostly taking already-saved audio and converting it to text.

The accuracy of the speech-to-text in AssemblyAI is the first valuable feature. I believe the most important aspect is not so much the functionality, but rather the performance and the low latency it has. In terms of features, the streaming part is particularly valuable.

The streaming functionality makes it easier for our team's work and impacts our daily work. Instead of having to record an entire audio file and then send it to the speech-to-text API, the audio can be transcribed in real time. This means if a person makes a mistake, they do not have to start the audio over from the beginning.

Another interesting feature of AssemblyAI is the functionality to recognize different speakers. For example, if I am recording a conversation or video call through AssemblyAI, the tool has the ability to recognize the different members or participants in the conversation. It can identify speaker one, speaker two, speaker three, and this can be very useful to analyze the conversation afterwards.

AssemblyAI has positively impacted my organization because our users started using the chatbot much more from the moment we implemented AssemblyAI in the streaming tool. Usage increased by 300%.

From my experience with AssemblyAI, I cannot think of any improvement that has not already been implemented.

I am not certain if it is already implemented, but something that could be interesting is that since AssemblyAI can identify different speakers through the frequency of the voice, it could maintain a memory of that same frequency so that in future conversations it can automatically identify which speaker is talking.

I started using AssemblyAI in 2023 for projects and side projects at university that needed speech-to-text transcription, converting speech to text.

AssemblyAI is stable in my daily use. I have not faced any outages or service failures.

AssemblyAI handles increases in data volume or users well.

I did not need to contact AssemblyAI's customer support.

Before using AssemblyAI, I used an open-source library for the speech-to-text component. This library had many errors, especially in the Portuguese language. The transcription to Portuguese was not as accurate and it also did not have the streaming functionality. These issues led me to switch to AssemblyAI.

My experience with AssemblyAI's pricing, setup costs, and licensing is pleasant. I was already familiar with the tool, so in this implementation and pricing part, I already had access and it was not a surprise. Implementation was easy.

My company does not have any commercial relationship with AssemblyAI other than being a customer.

I have not had a return on investment from using AssemblyAI yet.

I have not noticed any other measurable benefit, such as time savings or cost reduction, after implementing AssemblyAI.

Before choosing AssemblyAI, I evaluated other options such as Google's own speech-to-text tools. I also considered Gemini, but AssemblyAI had my preference since I already had more experience with it and it had lower latency.

My choice was based more on the experience I have had over the last few years.

I would advise people thinking about using AssemblyAI to try it first and test it, comparing it with other solutions, especially Google's Gemini text-to-speech. Try AssemblyAI first because it has a very generous free tier that allows building prototypes easily and seeing if it is the solution needed.

I believe I have covered everything about AssemblyAI. It is a very useful tool and the company itself appears to me to have always been evolving with new features since I started using it in 2023 until now. There is no stagnation; I can see that it has evolved. AssemblyAI is a reliable company that offers a generous free tier for prototyping, which is excellent for startups and side projects. I have had a pleasant overall experience and therefore my rating is 9 out of 10.

My main use case for AssemblyAI is to summarize content. Our company asks writers to write complete applications, tell complete stories, or narrate stories on our platform. By using AssemblyAI, we create text-to-speech and textual summarization features. Writers accumulate their work into scripts or summaries, and we use those summaries to produce novel summaries.

A specific example of how I use AssemblyAI in my workflow is that story writers write complete stories, and AssemblyAI summarizes the entire story so that readers can access the summary written by the authors. Additionally, we have used it for text-to-speech, and we have utilized AssemblyAI in our workflow for both of these cases.

The best features AssemblyAI offers are summarization and animations. The AI summarization feature stands out for me because we work with Pocket FM, and we utilize it for summarization of our novelists and novel writers' stories. We have also utilized AssemblyAI for different voice generation options.

AssemblyAI has positively impacted my organization because by using it, both our story narrators and story readers benefit greatly. We are able to successfully generate better summaries of stories and story chapters, and the text-to-voice feature is very helpful for our users.

Specific outcomes that show how AssemblyAI has helped my organization include increased engagement and positive user feedback for text-to-speech. User engagement has increased from 76% to 78% by using AssemblyAI.

One area where AssemblyAI can be improved is in summarization. Sometimes it does not fetch the correct words or identify the most important things for that chapter, so it needs to improve in this area.

I have been using AssemblyAI for two years.

AssemblyAI is stable in my experience.

The scalability of AssemblyAI for my organization is good as we have a high volume of scale right now, with around 25,000 requests, and it is performing well.

For customer support of AssemblyAI, I would rate it six to seven because sometimes they help us in a great way, but sometimes they do not help or take a long amount of time. AssemblyAI has a chatbot that helps a lot and assists us in accomplishing tasks, so I am satisfied with it.

Before using AssemblyAI, I previously utilized ChatGPT for summarization of text and chapters written by authors. This process took a large amount of time, which led us to switch to AssemblyAI.

On a scale of ten, I rate how easy it was to implement AssemblyAI in my environment as eight.

I integrate AssemblyAI with my existing systems and workflows by using the Python library.

For updates or maintenance of AssemblyAI in my setup, the updates do not come frequently, so whenever they come, we update the Python package library.

I have seen a return on investment from AssemblyAI through time saving. Before using other tools for summarization of writers' stories, it took a lot of time, but with AssemblyAI, the time to convert a story into a summary has decreased significantly.

Regarding my experience with pricing, setup cost, and licensing, I would like to know if it can be reduced somehow, but apart from that, licensing is good enough for us.

I did not evaluate other options before choosing AssemblyAI because I was referred to this platform by one of my seniors, so I utilized it directly and am happy with this software.

My advice for others looking into using AssemblyAI is to check the packages in different languages while integrating it into their use cases, so I recommend checking those availability options. I would rate this review overall as an eight.

I used AssemblyAI for a small task in my company where I had to create a chatbot, and my work was mostly converting speech to text.

My main use case for AssemblyAI was to create a voice-to-voice interactive chatbot. For the task of converting speech to text, I used AssemblyAI's API, which was quite good, with the best latency and a very good experience overall.

A specific example of how I used AssemblyAI in my chatbot project was in the pipeline that initially converted speech to text, then sent that text to an LLM for a response, then converted it back to speech again, sending it to the client's browser. I focused mainly on the speech-to-text conversion, which required AssemblyAI.

What stood out to me about the speech-to-text feature of AssemblyAI was the speed, accuracy, and ease of integration. All of these strong points contributed to a very good development experience while working with AssemblyAI.

AssemblyAI positively impacted my organization as we previously used Vapi for all voice-related chatbot tasks. Since I created the in-house chatbot using AssemblyAI and LLMs, our product became much cheaper, and we no longer need to rely on Vapi or Retell.

The main outcome since switching to AssemblyAI was cost savings. Although I cannot recall the exact amount we saved, I know we saved a fair amount using our chatbot.

I believe AssemblyAI needs to improve its filter for removing filler words. It works quite well and automatically removes terms such as 'um' and detects when I stop talking, so I think it is already up to the mark with latency and performance.

I wish AssemblyAI could improve its multilingual support, as it did not work well when I spoke in different languages. For instance, it works better in English than in Hindi or other languages.

No improvements are needed for AssemblyAI beyond the multilingual support I mentioned, as everything else seems quite good.

I have been working in my current field as an intern from January 2025 through April 2025.

I rate AssemblyAI a 10 because I had a specific use case and found it through a Google search by typing 'API for converting speech to text'. The experience I had integrating it was very easy, and I had no difficulties integrating AssemblyAI with my project. The output was excellent.

Regarding AssemblyAI's AI capabilities, I think the accuracy and reliability of output are up to the mark. Since I already gave it a 10, you can assume all my answers are positive.

If someone comes to me trying to build a chatbot, I recommend using AssemblyAI for the speech-to-text task. My recommendation alone carries weight.

I appreciate that AssemblyAI had a very good developer experience, although I do not remember all the specifics since it was a long time ago.

My overall review rating for AssemblyAI is 10.

The best features AssemblyAI offers are transcription and real-time transcriptions. The speed of real-time transcription stands out to me because it's 20 to 40% faster than the industry benchmark, so speed is definitely one of the pros of AssemblyAI.

AssemblyAI has positively impacted my organization by being a fundamental part of our main use flow, where our bot joins the meetings and transcribes them into text. Once the text is generated, it goes to our internal LLM to get culture scores, making it one of the main fundamental parts of our product.

AssemblyAI could be improved because when we have different accents on the same call, it usually fails, especially when we have American, Asian, and Latin American speakers on the same call, making the transcriptions a bit noisy.

The transcription quality of non-native English speakers should be improved. I choose nine out of ten because it's really good and fast, working well when there is an English speaker on the call, so the quality of the transcription is really good. Latency is almost zero, and it's 20 to 40% faster than the industry benchmarks. I only rate it as nine because it lacks accent detection and the quality for different accents.

My main use case for AssemblyAI is in my projects, such as an AI meeting assistant and AI interview platform, where I use AssemblyAI as a speech-to-text converter. I also use its stream option with AssemblyAI's webhook, and I record audio for speech-to-text conversion.

In the AI meeting assistant project, I specifically record meetings in my application and convert them into text. I then give the transcript to the AI, and it summarizes and makes quick notes. AssemblyAI's main function is to transcribe audio to text, and it supports multiple languages with good accuracy, although I think the accuracy decreases when I use the streaming option.

I mostly use AssemblyAI for transcription, whether it is recorded or streaming, as I am a developer who integrates AssemblyAI into my projects.

The best features of AssemblyAI are the stream support and multi-language support, along with speaker detection and SRT format. The stream support helped me in an AI interview platform where a candidate speaks, and I see the live transcription of their answers, which is helpful for candidates to identify if their words are correctly captured. Multi-language support is also beneficial in my meeting project, where I use languages such as Hindi and Marathi.

AssemblyAI has positively impacted my organization by being very helpful for audio transcription, saving time, and helping me find meeting notes.

I believe the streaming process accuracy should improve, and I also wish for a proper application for normal users, not just for developers, so they can easily convert audio.

I noticed that when I speak in Hindi, it sometimes writes in Urdu. AssemblyAI's accuracy is good, but it still needs improvement in some native languages, such as Marathi, while it works best in English.

I would like to reiterate the need for a proper application for normal users, not just developers, allowing them to convert audio to text and analyze it.

I noticed that when I try to upload a large file on AssemblyAI, it sometimes gives an error such as session timeout, and it has a limit of 500 MB file size.

I have been using AssemblyAI for two to three months.

AssemblyAI is stable in my experience.

Before choosing AssemblyAI, I evaluated it alongside Whisper, but I feel AssemblyAI is better compared to Deepgram.

Using AssemblyAI has saved me time and money.

Regarding my experience with pricing, setup cost, and licensing, I purchase from multiple platforms such as cloud, GCP, AWS, and other providers including ChatGPT. My competitor Deepgram offers lower cost but does not support my native language.

Before choosing AssemblyAI, I evaluated it alongside Whisper, but I feel AssemblyAI is better compared to Deepgram.

My advice to others looking into using AssemblyAI is to use the stream support with proper documentation. I would rate this product an 8 out of 10.

I use AssemblyAI only with audio files, not for real-time transcription. I mainly use only US English, and I have not tried other languages. I upload audio files through AssemblyAI API, and they provide the transcription script with speaker identification and the dialogues.

The main features I appreciate in AssemblyAI are that it provides better accuracy compared to other transcription services, with clear grammar and no errors in spelling mistakes or grammatical mistakes, delivering clear transcription.

The primary benefit I receive from their product is much more accurate transcription. First, it is a very affordable service, and second, the accuracy is much better compared to other services such as Deepgram or AWS transcription services, which are the main benefits. Third, the speaker identification capability is better.

A few drawbacks I observed in the speaker identification are that in some videos where text and names appear on the video frames, AssemblyAI does not identify the actual speaker name, instead providing generic names such as Speaker A, Speaker B, Speaker C, or Speaker X, Y, Z.

AssemblyAI does not identify the real speaker in some audio or video files, just sending Speaker A, Speaker B, or Speaker C. They are not easily identifying speakers in some instances.

AssemblyAI does not provide a cloud service; I simply upload the audio file to the API, and they store it somewhere internally to send me the transcription text.

For additional functions, the API does not provide video uploading functionality, and I need to convert video to audio first before uploading it to AssemblyAI.

I have been working with AssemblyAI for approximately one year.

AssemblyAI should respond more quickly because when I post a ticket, they take too much time to respond to it.

I did not continue working with Deepgram after trying it, but I recently started using AssemblyAI because Deepgram does not provide accurate transcription. I chose AssemblyAI because I did not use Deepgram again.

I only need to create an account on AssemblyAI, and initially, they provide some credits for transcription, which is enough initially. However, if usage increases, I can purchase a subscription from there.

I think the price for the product is a seven.

I can compare AssemblyAI with Deepgram. I would choose only AssemblyAI instead of Deepgram when comparing both products. The main reason I chose it is that it is far better compared to Deepgram regarding speaker identification, the clear verbatim process, and the time-stamp process, providing accurate time-stamping and the dialogues.

If I compare AssemblyAI with other services such as Gameloop, ChatAI, and Deepgram, the accuracy is far better, always maintaining the grammar and providing good, accurate text for audio or video files.

The AssemblyAI noise filtering feature exists, but I did not use that feature. I use the existing API where I upload the audio to AssemblyAI, and after a few seconds or minutes, I continuously check if the transcription is done. Once it is done, I pass the transcription text into a file and generate an SRT file, a text file, and a doc file.

It works fine with different accents.

I rate this product an overall 8 out of 10.

My main use case for AssemblyAI was for a voice-based calling system that we created.

A specific example of how I used AssemblyAI in my voice-based calling system is that it was used for the text-to-speech part and also speech-to-text. For speech-to-text, I don't remember if we also used something else, but for text-to-speech, we definitely relied on AssemblyAI because the voices were significantly better.

I don't have anything else to add about my use case with AssemblyAI.

The best features AssemblyAI offers include a large selection of voices and an intuitive dashboard where we could customize and tweak things according to what we needed.

When I mention customizing and tweaking things, it was a mix of all aspects, and I could also change the AI model behind it.

AssemblyAI has impacted my organization positively, not in the current organization, but in the previous organization. Because of AssemblyAI, we were able to envision something that could speak naturally to mass consumers.

That natural-sounding speech impacted our customers because users couldn't tell if it was AI, and even if they could, they could still have a natural conversation.

AssemblyAI can be improved in several areas. I don't remember if AssemblyAI had a webhook, but if they have one, that's beneficial. The biggest improvement I think was the lack of background sound. If AssemblyAI doesn't have this feature anymore, but suppose we have a use case where a customer support representative is talking to an actual customer, for it to actually sound natural, perhaps a dim office workspace background would be nice. This would give it a much more natural feel instead of a very silent background, which feels very artificial.

Regarding AssemblyAI's AI capabilities, I have no idea about its governance and security. I would really like to know if it is compliant with Saudi PDPL. A feature I would want AssemblyAI to have is to cater to the MENA region and the Middle East market. There's a law in Saudi called PDPL, Personal Data Privacy Law, and that dictates that all the data should remain within the Kingdom or within GCC and all the processing should happen there. If AssemblyAI could offer some feature or some way to ensure that the hosting is being done within maybe GCP Dammam or AWS in the GCC, or the cloud providers that reside here, and the physical data centers that are being used are of the GCC for our entire processing, that would make it so much more feasible to be used within organizations in Saudi.

Regarding AssemblyAI's AI capabilities, I can't think much about its accuracy and reliability of output because I think the AI is of a third party that is connected. I don't know if AssemblyAI has any LLM that does the reasoning. So I don't think it's fair to say it's AssemblyAI's capability.

I used AssemblyAI for one and a half years, but I haven't been using it for a year now.

I would say to others looking into using AssemblyAI that it's a good start and it's very easy. They should start by first using the dashboard and trying it out instead of jumping into integrations to see if it best fits their use case. I would rate my overall experience with AssemblyAI an 8 out of 10.