--- schema: 1 type: video title: 'While my guitar gently speaks — Todd Fisher, Philo Ventures' description: >- Someone in the audience asked the guitar what reality is, and the guitar answered. Todd Fisher's build routes a microphone through speech recognition into a local model and pushes the reply back out through the strings, which is the most recent step in a project that began with a much simpler question: how hard could it be to make a guitar speak? The lineage he draws runs from a pickup and an amplifier, through stomp boxes, to Peter Frampton sending guitar sound down a physical hose into his mouth. His own version is a plugin built with JUCE that drops into a DAW like any other effect. Getting it to say one word was straightforward. Getting it to say several meant slicing synthesized speech into words automatically, and that turned out to be the hard part. Energy gap segmentation cuts wherever the signal falls toward silence, which fails because running speech often has no silence between words at all. A sonority peak syllabifier looks for vowels instead. Combining the two got close enough that he finished by dragging segment boundaries by hand. Singing needed a different stack again: the YIN algorithm to pull a fundamental frequency off each fretted note, a synthesized tone shaped by an envelope, then a vocoder, with pitch shifted samples from an open singing dataset baked ahead of time because the processing is far too heavy to run live. He also declines to play the song his title alludes to, on the grounds that this recording was going online. Speaker info: - https://www.linkedin.com/in/todd-b-fisher Timestamps: 0:00 - Live performances that stayed with him 2:44 - The guitar's evolution, up to the talk box 4:24 - A Halloween project on a garage door 6:06 - Building it as a JUCE plugin, and saying one word 7:50 - Slicing speech into words, and why that is hard 10:23 - Pitch detection with the YIN algorithm 11:15 - Synthesis, vocoder, and jamming with it 13:28 - A guitar that answers questions from the room 16:01 - Pitch shifted samples, and getting closer to singing 17:51 - Go build your passion project canonical: >- https://liquidgold.cc/channel/aidotengineer/explore/video/while-my-guitar-gently-speaks-todd-fisher-philo-ven-7963c043d527 markdown: >- https://liquidgold.cc/channel/aidotengineer/video/while-my-guitar-gently-speaks-todd-fisher-philo-ven-7963c043d527.md llms: null parent: 'https://liquidgold.cc/channel/aidotengineer.md' channel: aidotengineer updated: '2026-10-24T05:13:36.685Z' language: null children: [] videoId: E_Txocq-Lrw durationSeconds: 1115 publishedAt: '2026-08-18T15:30:23.000Z' searchChunks: - ordinal: 0 text: null startSeconds: 23 endSeconds: 170 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=23' heldReason: rights - ordinal: 1 text: null startSeconds: 23 endSeconds: 170 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=23' heldReason: rights - ordinal: 2 text: null startSeconds: 181 endSeconds: 267 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=181' heldReason: rights - ordinal: 3 text: null startSeconds: 181 endSeconds: 267 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=181' heldReason: rights - ordinal: 4 text: null startSeconds: 181 endSeconds: 267 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=181' heldReason: rights - ordinal: 5 text: null startSeconds: 282 endSeconds: 374 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=282' heldReason: rights - ordinal: 6 text: null startSeconds: 282 endSeconds: 374 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=282' heldReason: rights - ordinal: 7 text: null startSeconds: 385 endSeconds: 470 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=385' heldReason: rights - ordinal: 8 text: null startSeconds: 385 endSeconds: 470 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=385' heldReason: rights - ordinal: 9 text: null startSeconds: 480 endSeconds: 632 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=480' heldReason: rights - ordinal: 10 text: null startSeconds: 480 endSeconds: 632 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=480' heldReason: rights - ordinal: 11 text: null startSeconds: 632 endSeconds: 643 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=632' heldReason: rights - ordinal: 12 text: null startSeconds: 643 endSeconds: 675 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=643' heldReason: rights - ordinal: 13 text: null startSeconds: 643 endSeconds: 675 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=643' heldReason: rights - ordinal: 14 text: null startSeconds: 821 endSeconds: 965 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=821' heldReason: rights - ordinal: 15 text: null startSeconds: 821 endSeconds: 965 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=821' heldReason: rights - ordinal: 16 text: null startSeconds: 965 endSeconds: 976 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=965' heldReason: rights - ordinal: 17 text: null startSeconds: 976 endSeconds: 1079 watch: 'https://liquidgold.cc/watch/E_Txocq-Lrw?t=976' heldReason: rights --- # While my guitar gently speaks — Todd Fisher, Philo Ventures Someone in the audience asked the guitar what reality is, and the guitar answered. Todd Fisher's build routes a microphone through speech recognition into a local model and pushes the reply back out through the strings, which is the most recent step in a project that began with a much simpler question: how hard could it be to make a guitar speak? The lineage he draws runs from a pickup and an amplifier, through stomp boxes, to Peter Frampton sending guitar sound down a physical hose into his mouth. His own version is a plugin built with JUCE that drops into a DAW like any other effect. Getting it to say one word was straightforward. Getting it to say several meant slicing synthesized speech into words automatically, and that turned out to be the hard part. Energy gap segmentation cuts wherever the signal falls toward silence, which fails because running speech often has no silence between words at all. A sonority peak syllabifier looks for vowels instead. Combining the two got close enough that he finished by dragging segment boundaries by hand. Singing needed a different stack again: the YIN algorithm to pull a fundamental frequency off each fretted note, a synthesized tone shaped by an envelope, then a vocoder, with pitch shifted samples from an open singing dataset baked ahead of time because the processing is far too heavy to run live. He also declines to play the song his title alludes to, on the grounds that this recording was going online. Speaker info: - https://www.linkedin.com/in/todd-b-fisher Timestamps: 0:00 - Live performances that stayed with him 2:44 - The guitar's evolution, up to the talk box 4:24 - A Halloween project on a garage door 6:06 - Building it as a JUCE plugin, and saying one word 7:50 - Slicing speech into words, and why that is hard 10:23 - Pitch detection with the YIN algorithm 11:15 - Synthesis, vocoder, and jamming with it 13:28 - A guitar that answers questions from the room 16:01 - Pitch shifted samples, and getting closer to singing 17:51 - Go build your passion project Canonical: https://liquidgold.cc/channel/aidotengineer/explore/video/while-my-guitar-gently-speaks-todd-fisher-philo-ven-7963c043d527 · Channel: [AI Engineer](https://liquidgold.cc/channel/aidotengineer.md) > Search chunk text is held until public-source provenance is verified. ## Document chunks > Missing: no stored document chunks are available.