Yes, I've seen that the "X" API's use has skyrocketed in price. Maybe Odysee's and Rumble's are still economically accessible. Or your service could be financially supported by INLEO, as part of the platform's native features.
You are viewing a single comment's thread from:
It's not accessing the various services that is expensive. It the actual transcription of the audio, i.e. running the voice to text software.
If I run it on my own hardware, 2 hours of taking takes about 20 minutes with my RTX4090 (high-end, consumer grade graphics card). This is how I currently transcribe x spaces - I fetch the link from the site, download it to my computer, run the transcription software (Whisper AI) on the file and then I do a summarization of the transcript using one of the AI model providers (currently Anthropics Claude)
If I were to do the transcription via a text to audio API (like openai), it would get very expensive very quickly. So in other words, if there is no transcript, no summary. This is why I've suggested incentivized video transcription on spknetwork, so that people can volunteer their own hardware to transcribe videos. I've promised them that the moment there are transcripts on 3speak, I'll make the ai summarizer work for 3speak videos as well.