Swahil Audio2English Answer
|94%
Jacaranda + GPT5 + Google Trans
|100%
|49%
GPT4o realtime beats GPT-realtime 31%
| |Swahili Audio2SwahiliAudio
Mean in seconds
| 6.3 | 5.99 | 6.48 | ### **Explore the detailed evaluations:** **KINYARWANDA** {% embed url="
{% @arcade/embed flowId="tU2V5JDJFls3yfWbCaqJ" url="
| Nurse Julie for MEXA: Voice conversations to patient records | /files/zQ1TLC1dIn88BqMaK9KM | https://gooey.ai/copilot/nurse-julie-for-mexa-voice-conversations-to-patient-records-qa7ym21pera0/ |
| Workplace Mental Health Bot | /files/8AqdxfrIfhAwucMbMS0R | https://gooey.ai/copilot/workplace-mental-health-bot-rckpkrusqhf0/ |
| Transcribe Patient History | /files/1KjJM9kbdSMzgD99pOeL | https://gooey.ai/speech/whisper-large-v2-openai-speech-recognition-and-translation-okinsy63gnwl/ |
Demo of our Web Widget with Sources
An artwork depicting the nuance of database software. Image generated with AI.
The architecture diagram of Gooey.AI. Watch our explainer video to learn more.
| How to use Gooey.AI functions? | https://docs.gooey.ai/guides/how-to-use-gooey-functions | /files/cGJknqvKaL2exH0zaOvN |
EXAMPLE: Fetch WEATHER API | https://gooey.ai/functions/current-weather-rxmquy60p1vq/ | /files/EjpuuFtzzW7uwB4jwQ1t |
EXAMPLE: Connect Fetch API with GOOEY.AI LLM Generator | https://gooey.ai/compare-large-language-models/functions-make-a-haiku-with-iss-coordinates-k4vuehh6hhvo/ | /files/nzBy9PMfGIWXVTbRLdCW |
POP-UP WIDGET
INLINE WIDGET
| GUIDE: Agent Web Integration | https://docs.gooey.ai/guides/copilot/deploy-to-web | /files/UQCFTw0JXqZyAWJlqWsv |
| GUIDE: Agent WhatsApp Integration | https://docs.gooey.ai/guides/copilot/deploy-on-whatsapp | /files/KPILIYJxcgKg8c6JeDcg |
| BOOK A DEMO! | https://www.help.gooey.ai/contact#book-demo | /files/novXSAfZ8SgXWbmaia1n |
| Language | Partner Data Set Link | Top Performing Model | Runner Up | Link to Evaluation |
|---|---|---|---|---|
| Hindi | ARTPARK (IISc) | GPT-4o | ElevenLabs Scribe v1 | Hindi Bulk Run |
| Swahili | Gates Foundation | Jacaranda + GPT5 | Jacaranda + GPT5 + Google MT | Swahili Bulk Run |
| Kikuyu | Gates Foundation | Akera+GPT5 & Akera+Gemini2.5pro+GoogleMT | Akera+Gemini2.5pro | Kikuyu Bulk Run |
| Chichewa | Opportunity | Seamless M4T v1 + Google Translate | MMS | Chichewa Bulk Run |
| Kinyarwanda (*new) | Gates Foundation | Kinyarwanda (Mbaza+GPT5) + Kinyarwanda (Mbaza+Gemini2.5pro) + Kinyarwanda (Mbaza+GPT5+GoogleMT) | Sunbird+GPT5 | Kinyarwanda Bulk Run |
| Magahi | Looking for Collaborator/Partner | |||
| Luo/Dhuluo | Looking for Collaborator/Partner | |||
| Maithili | Looking for Collaborator/Partner |
š GUIDEHow to create language evaluation for ASR? | /files/EUdE0ZYVMYls2jfQkAcZ | /spaces/5BFP5RUm6rTLXk8wUSTf/pages/Dv4ARYu6p13xP3yixDMg |
š£ļø Hindi ASR EvaluationGet started with a pre-filled example | /files/nuK6Egv9ItSNLNGgvZ1v | https://gooey.ai/bulk/compare-hindi-speech-recognition-hkgs8120p11t/ |
Screenshot from our Bulk and Evaluation tool, showing the difference in quality of answers from the copilot.
| Start building your own Agent | https://gooey.ai/copilot |
| Check our Agent Guide | https://gooey.ai/docs/guides/copilot |
| Book a Demo | https://www.help.gooey.ai/contact |
# The 2023 Gooey.AI Recap \- A Letter from the Gooey.AI Founders Hello world, We started Gooey.AI in September 2022 as an idea to spread AI innovation to everyone. We believed then (as we do now) that most of us will become **AI prompt writers** - cajoling it to create useful artifacts for us - and **API stitchers** - connecting AIs to other datasets, AIs or tools. We thought the next logical computing metaphor would be **AI workflows** - small recipes of AI prompts that run on LLMs and interact with APIs. We believed Gooey.AI could be useful to every person and organization if we could make it simple and fun to learn from each otherās AI recipes and then run, tweak and re-share them. We felt AI workflows should be built from an ever-expanding foundation of private and open-source AI innovations; people should be able to compare, hot-swap and integrate the latest innovations with minimal effort to facilitate maximum experimentation and innovation. Itās been incredibly gratifying to see this vision come to life in the last year. * **š 243,345 people** ran over **1.4M Gooey.AI workflows** in 2023. * š Gooey.AI ranks among the top 5 Google results for [AI Animation](https://gooey.ai/animation-generator/), [AI QR Code](https://gooey.ai/qr-code/), [Lipsync](https://gooey.ai/lipsync-maker/) and [AI Prompt Image Editor](https://gooey.ai/ai-photo-editor/). * š¢ We released our [open-source vision](https://gooey.ai/blog/gooey.ais-open-source-vision), [explainer video](https://gooey.ai/docs/#explainer-video) and [20+ forkable AI recipes](https://gooey.ai/explore). * š We launched our [/blog](https://gooey.ai/blog) and [/docs](https://gooey.ai/docs) subsites for Gooey.AI updates and guides. * š¤Æ We published our [guide to thriving in the All-AI agent Internet](https://gooey.ai/blog/the-genai-marketing-disruption-and-how-gooeyai-can-help) (including tips to craft [beautiful QR Codes](https://gooey.ai/qr-code/) and [zeitgeisty āPeople Also Askā answers](https://gooey.ai/related-qna-maker/) to help your site rank better). * š± Our partner DigitalGreen showcased [Farmer.CHAT](https://www.help.gooey.ai/farmerchat) at the UN General Assemblyās [Science Panel](https://webtv.un.org/en/asset/k1v/k1vzgefyvn?_gl=1*1njkmi4*_ga*MTU1Mzc1NTU4NC4xNjgwODk1OTM2*_ga_TK9BQL5X7Z*MTY4MTI2Mjc2MC40LjEuMTY4MTI2MjgwMC4wLjAuMA\&kalturaStartTime=7697\&kalturaStartTime=7701\&kalturaStartTime=7701) - a multilingual WhatsApp AI agent for Indian, Ethiopian and Kenyan farmers. 1000+ farmers asked Farmer.CHAT over 35,000 questions and itās been heartening to see other organizations re-use and extend the [Farmer.CHAT agent](https://gooey.ai/bots/?example_id=nuwsqmzp) recipe to more people and places. * šŖš¾ Our [bulk](https://gooey.ai/bulk) and [eval](https://gooey.ai/eval/examples/) tools highlighted how well (or poorly) the top speech recognition tools from OpenAI, Bhashini, Google, Meta and Azure understand low-resource languages. Bring your own audio clips to compare how well they work for you or use the [eval tool](https://gooey.ai/eval/examples/) to compare and evaluate any collection of Gooey.AI workflows. * šø Our AI agent[workflow](https://gooey.ai/docs/guides/copilot) now supports vision (so you can use a photo, audio clip or text as input), indexes complex documents (with tables), creates synthetic data (like FAQs) from documents and videos, [understands speech and translates](https://gooey.ai/speech) hundreds of languages, [speaks in custom voices](https://gooey.ai/compare-text-to-speech-engines/) and [lipsyncs videos](https://gooey.ai/lipsync-maker), analyzes and categorizes conversations, can be tested via bulk evaluations and deploys with WhatsApp, Facebook, Slack, Instagram and more. Itās pretty amazing. * š¾ We launched Saved Workflows on Dec 25 and already hundreds of people are saving workflows to deploy in their own apps and share back to the community. š®And to give you a taste of whatās to come in 2024: * **Organizations** so multiple people can collaborate on a workflow and teams can give every employee access to the best of the GenAI universe * **Trending AI workflows,** organized by industry * **SHIPPED! Profiles** so you can find the best Gooey.AI creators (or become one!) * [**SHIPPED**](/from-bland-to-brilliant-ai-agents-new-agent-features-for-april-2024)**! Gemini Pro 1.5, LLaMA3, Claude3, and Mistral support** for [LLMs](https://gooey.ai/llm) and [Copilots](https://gooey.ai/copilot/) (in addition to our existing support for GPT4-turbo, GPT-Vision, Palm2 and LLaMA2) * **SHIPPED! Streaming support** over WhatsApp, Slack and web-based agents As we close this year, we give thanks to our incredible [team](https://gooey.ai/team), investors (especially [Bruce Jaffe](https://www.linkedin.com/in/brucejaffe/) of [J4](https://www.j4.ventures/) & [Blume](https://blume.vc/)), advisors (especially [Vir Kashyap](https://www.linkedin.com/in/vkash/)), partners (thank you [People+.ai](https://peopleplus.ai/), [DigitalGreen](https://www.digitalgreen.org/), [GenAISolutions](http://genaisolutions/) & [ArtPark.in](https://artpark.in/)), families, clients and users. We wish for peace in the year ahead and we hope love, creativity and innovation brings prosperity for us all. Sincerely, Sean Blagsvedt, Archana Prasad and Dev Aggarwal Founders, Gooey.AI PS. Like the masthead picture? [Hereās the workflow](https://gooey.ai/compare-ai-image-generators/?run_id=5vp013up\&uid=cs66PB2T61Tnkg2CPN1btRP9WEd2) that made it. # The GenAI Marketing Disruption & How GooeyAI Can Help GenAI changes how every brand & website gets found. Here's how the tech giants will respond + how our tools can help the rest of us. ### The All-Bot Internet + How the Search Engines (and LLMs) will respond. LLMs such as GPT5 and Gemini are amazing, somewhat terrifying and absolutely disruptive. They have hacked human language and can write optimized content about anything. Hence, whatās a poor search engine like Google or Bing to do when presented the task of showing (or creating an answer from) the most relevant, up-to-date webpages when asked a question? As AI agent-optimized content spreads across more of the web (see [jasper.ai](https://www.jasper.ai/)), the engines will rely less and less on the *content* of a page - the words, images and videos - because that content can be fairly easily created and optimized by agents. Hence, they will rely more and more on the SOCIAL SIGNALS of the page and its website domain. Was it from a credible site e.g. the New York Times? How many other high-authority sites link to this site? How is the site reviewed on Google Maps? How do users behave on the site and its page? Do they engage deeply and spend significant time there? Are there 1000s or millions of other pages discussing this site (and hence influencing the LLM training data itself)? Hence, great content for a page or site is now table-stakes - a required but not sufficient condition for Google, ChatGPT, Claude, etc to consider and link back to your site. Once you create that incredible, relevant content (and your competitors likely already have), the real work begins, moving the SOCIAL SIGNALs. ### The Social Signals 1. **Real-Life Reviews** e.g. Google Maps. Real reviews from real individuals (usually created by people at a particular place on their phones - not on desktops) are a difficult-to-fake signal of true value, especially for web searches that are tied to geographical places. Google, Meta and Bing will increasingly lean into these signals as *content* is increasingly gamed by AI agents. 2. **Time on site aka engagement**. Have you noticed that every food recipe site has turned into an [infinite scrolling timesuck](https://www.loveandlemons.com/how-to-cook-rice/) to list 3 ingredients? Wonder why? Google is measuring the time between when you (and millions of others) click on a link and later tap āBackā to the search results. If that time is short, Google infers that the link it offered up was NOT that useful and hence, lowers the pageās rank in the future. If Back was rarely clicked or the timespan is long - e.g. you spent 5 minutes scrolling through those beautifully lit videos on how to mix flour, butter and sugar - Google infers that the link satisfied your query and effectively ups the pageās rank. Iād venture that if a ChatGPT or Gemini question employed a real-time web search and the user seemed to like or dislike the answer (as evidenced by their subsequent chat messages), that signal will inform the pageās future rank against that query as well. But more on this in a future post. 3. **Virality via Social links** - similarly if millions of people are posting links on social media back to your site e.g. sharing Pinterest board links, it signals to the engines that your site is making something worth sharing and hence valuable. 4. **Press links aka high-authority backlinks** - if lots of high-quality news sites are discussing your brand or product and are linking back to your site, Google notices. PR and Press are hard work but by extension, they are hard to fake with agents. ### Gooey.AI tools to fight back #### Beautiful QR Codes for Real Life Engagement If you want your storeās rank in Google Maps AND Google Search to increase, your customers must give you positive reviews on Google. If you have physical locations, ideally those reviews are given on their phone and when they are present at your business. Hence, you have to increase their motivation to positively review you and reduce the friction of doing so. Scanning QR codes have become the most common way to induce this action and weāve discovered that if you make those QR codes beautiful and interesting, scan rates increase. Art-based QR Codes represent a truly novel new form of aesthetics - the sort of thing that humans really couldnāt easily make - and hence, weāve discovered thereās a curiosity among observers to determine if they scan. Furthermore, beautiful QR Codes enable user appreciation of the QR code itself which increases motivation to scan it. Weāve built [the best QR maker in existence](https://gooey.ai/qr-code/) right now - one that leans towards scan reliability while blending the QR Code seamlessly with gorgeous AI generated visuals. Weāve got tons of examples and premium analytics available too.   #### Higher Time-on-site & Engagement with AI agents and Quizzes Getting users to stick around is hard but in our work with other clients, weāve discovered a few insights. 1. **Agents are a thing.** [Character.AI](https://beta.character.ai/) has the highest engagement of any app on the AppStore. OpenAI just launched GPTs as a competitive response. We've built [Gooey.AI Copilot](https://gooey.ai/copilot.) - [the most advanced bot maker](https://gooey.ai/copilot/) on earth that works across every major LLM (and speech recognition engine) and connects to SMS/Voice, WhatsApp, Facebook, Slack or your own site. We also offer our services to select clients to build best-in-industry agents like [Farmer.CHAT](https://www.help.gooey.ai/farmerchat) which was demoed at the [UN General Assembly](https://www.help.gooey.ai/farmerchat).  2. **AI Content that Engages** If people really are into a concept, they competitively want to prove their level of in-group knowledge. The old [InStyle magazine quiz format](https://www.instyle.com/celebrity-style-doppelganger-quiz-8549598) is still fun & engaging and if you do well, itās the type of achievement you want to share with your friends (see point below). Weāve found that [AI-generated quizzes](https://gooey.ai/google-gpt/?example_id=h0d7v0pi) - when combined with real-time content and appropriate safety features - significantly drive time on site.  ### **Virality via AI-Generated Content** How do you convince most of your users to post a link back to your site? If you can achieve this, you increase organic word-of-mouth marketing while also, signaling to the search that youāve built something people value. But how? We believe the answer lies in creating personalized artifacts people want to share with their friends. examples include:
AI ART QR Codes on Labels
āEvery business will become an AI business.ā
Investing in AI to provide better & cheaper and scalable services in healthcare
Example workflows are publicly viewable, runnable and forkable on Gooey.AI
A userās workflow may be made public or kept private.
Future: Can be downloaded locally
| | Workflow Orchestration RunTime | The code required to execute a workflow (e.g. call an LLM with a prompt and send the result to another service) |Future: Open sourced on GitHub
Future: Hostable as a Docker container
| | AI Model Cluster | The collection of 50-100 open source AI models that must run on fast GPUs (e.g. LLaMA2, Bhasini Speech Recognition) |Future: Open sourced on GitHub
Future: Hostable as a Docker container
| | Analytics DB | Stores history of workflow runs, /agent conversation history, visualizes data |Future: Open sourced on GitHub
Future: Included with Docker container
| | Gooey.AI Workflow Directory | The public collection of workflows and discovery interface | Hosted on Gooey.AI and not expected to be open sourced. | | Private AI Services |The platform aggregates many paid AI API services such as:
- Private AI APIs: OpenAI, Google, Azure, AWS, Replicate, uberduck
- Communication platforms: Facebook/Instagram/WhatsApp, Slack
- Other APIs: SearchSERP, Contact Lookup
Large GPU cluster to host workflows, runtime, AI model cluster + analytics DB
Optional private AI keys.
| To Gooey: No per run fees. Optional consulting/support. A license fee may be required for large, private orgs. | | Hybrid | Organizations who to want privately manage their AI workflows but still want to experiment with the latest models. |Moderately capable server to host the workflow runtime.
Private AI keys.
| To Gooey: Per run fees whenever a workflow requires a private AI API call for which the org doesnāt have a key. Per API call fee to use our cloud AI model cluster. | ### Outcomes * Development orgs can prototype, test, measure impact and iterate faster at much lower cost * The latest AI innovations from the ecosystem get deployed in real systems faster * Every org gets to reuse and tweak the best performing AI interventions of others * Greater sharing of tech and AI components such that new innovations are published as components in Gooey and then adopted across many of its ecosystem users. * Shared measurement infrastructure * Insights and data on how research based knowledge is being translated by AI systems into advisory thatās being adopted or not. ### How this approach changes how large funders such as BMGF + EkStep invest in tech projects: Example: **An organization wants to create a local-language IVR front-end for ChatGPT.** Imagine a Kenyan NGO proposes to BMGF a great potential innovation - allow non-literate, non-smartphone users to call a phone number, ask any question in their local language and receive back an audio answer, leveraging the OpenAI GPT LLMās knowledge base via the API that powers ChatGPT.  Under the current funding system, this organization would have to build (or use private models) to create all the components. They would: 1. Get their own API keys to various LLMs like OpenAI 2. Determine which speech recognition and synthesis models worked best for their users and then host that infrastructure on their own GPUs. 3. Build a robust connections among LLMs, their speech recognition/synthesis components and whatever IVR system they choose 4. Build feedback and analysis systems to determine usage and retention patterns and do cohort analysis. 5. The code would likely not be structured for easy reuse by other organizations. If this were a Gooey ecosystem funded project, the execution would significantly differ. 1. Faster time to validation. By re-using components already in Gooey, the NGO can perform large-scale usability testing faster to determine if a wider roll out is merited. E.g. They could modify the LLM script of