The best of the AI tools (July 2023)Recently, I went down the rabbit hole of tools for the AI hackathon we recently had at the company I work at.
I tested over 20+ tools and I believe there are some early winners across several categories.
This is written as a raw review & stream of consciousness
Transcription + AI Summaries
Pros: Freemium model, super accurate, Integrates to the big-3 video platforms, also transcribes files async
Cons: gets expensive quick if you’re using it often, chrome extension
Pros: 1 hr free/month, Works well with large files, has an editor + allows you to note who’s speaking (and it saves their voice tag), great for subtitling videos (async), has studio-quality capabilities
Cons: Doesn’t do live transcription from video calls
Pros: Free, integrated directly into Zoom, the “highlights” feature is pretty neat (reminds me of clipping in the gaming/streaming space), provides full transcription, Their team edition provides some interest insights
Cons: Only works with the big 3 apps (pro/con),
Grain
https://grain.com/
Pros: integrates to zoom, g-meet, teams, FREE, and super accurate, + slack integration,
Cons: walled-garden (meaning you have to use those video apps), summaries are meh
Pros: AI summary is one of the best I’ve seen + great popup reminding you to start the summary process
Cons: AI features are all paid ($10/mo), requires you to play audio through macbook speakers (out loud), sometimes the process doesn’t “finish”, still in “Beta”, Doesn’t give you the full transcript only the summary
Video-esque + Audio-esque tools
Framedrop.gg - auto-clips for twitch streamers (highlights tool)
Pros: solves a common problem for streamers (free). It takes a fair amount of time to generate good clips to share out on social/video platforms from your stream — this does it automatically
Cons: It says it will always be free (but that is a known trap)
Play.ht - text to speech, voice cloning, AI voces
Pros: One of the best tools I used this year, voice cloning is super accurate, voice-to-voice :+1::skin-tone-5:, text-to-voice :+1::skin-tone-5:, their editor is slick as well, has an API
Cons: their mp4 outputs don’t work well on Apple products (no audio)
Mubert.com - generative music app
Pros: quickly get to royalty free music that’s “unique”, free, full length songs from a prompt, has an API
Cons: often gets the genre wrong, songs feel cheap at times.
Murf.ai - text to speech, voice cleaning, real voices
Review: this is half as good as Play.ht (above) — it does everything just “okay”
Voicemod - real-time voice change (AI)
Pros: soundboard + AI voices + “real voices”, works natively as an audio-input device when running locally, could be great for the future of faceless video content
Cons: expensive (paid upfront), creating a “voice” from scratch isn’t that great, the app is very large on the machine (700MB)
Synthesia - ai human voice explainer videos
Pros: it works? The quality of the voicedub to lipsync is shaky at times, great for kickstarting ideas, with about 30 minutes of work it can save hours of video editing + video recording a real human + removes the need to write a script
Cons: $22/month
Social + Productivity Tools
Postwise AI - AI tweet bot that writes tweets in your voice
Pros: free (for now), the tweets are well written and actually “sounds like me”, indistinguishable from my “real tweets”, scheduled tweet feature, great if you’re trying to “organically” grow your social footprint
Cons: too many emails from their team, the first tier of the paid option is too expensive $37/month
ChatGPT
Pros: free-ish, great at composing articles, blogs, stories, fiction, etc, APIs are great!, great as a productivity assistant to help generate ideas
Cons: bad at “facts” that have nuanced details, bad at current events
Images: MidJourney vs StableDiffusion vs DreamStudio
Pros: MidJourney is clearly the winner here, however it’s major flaw is that it only runs in Discord
Cons: StableDiffusion & Dreamstudio are just “okay” — but it’s often clear that the image has the AI glitch/blur in the image making it appear unreal/unrealistic