AI Filmmaking
Expanding creative possibilities with AI-enhanced filmmaking.
AI Filmmaking: From Creative Vision to Execution
As a filmmaker, I approach AI filmmaking as an extension of the traditional creative process, grounded in more than 20 years of experience in directing, cinematography, editing and audiovisual production. AI does not replace filmmaking craft or vision; it expands the range of ideas, worlds and production challenges that can be brought to the screen.
AI filmmaking is not simply the use of one model to generate an image or a video clip. It can involve an AI pipeline: a structured and partly automated workflow that moves from scripts, original footage, references, brand materials and prompts through preparation, generation and refinement, and into editing, sound, colour, mastering and delivery. The tools and models are selected and connected around the needs of the project, rather than allowing one platform to dictate the result.
The AI filmmaker turns a creative vision into an executable process: defining the concept and visual language, directing iterations, judging the outputs and integrating different elements into a coherent finished work. As powerful tools become accessible to almost everyone, the real differentiator is no longer access to technology or spectacle alone, but the originality of the idea, human sensitivity, creative judgement and the ability to turn a concept into a coherent experience. Each project remains human-led: the filmmaker is both the creative author and the executor of the concept.
Anyone can now build a universe. The real question is: who will care enough to step inside yours, and why?
In an ocean of content, attention is earned not through spectacle alone, but through vision, meaning and execution.
AI Video and Hybrid Filmmaking
These projects focus on the image: generative video, character animation, synthetic performances, AI-assisted visual effects, continuity reconstruction and hybrid workflows combining generated material with live-action production.
The Return — AI-Generated Documentary Proof of Concept
“Can an AI-generated production retain the quiet observation and emotional credibility of documentary cinema?”
An independent proof of concept developed during the exploratory stage of an unrealised production. Set in a fictional rural community in Portugal’s interior, the film follows an elderly man through the landscape, village streets and rhythms of everyday life.
AI-generated imagery and animation were subsequently shaped through human-led creative direction, character development, visual continuity, traditional editing and post-production. Selected portions of the generated clips were restructured and refined to mitigate visual errors, establish continuity and create the rhythm of the narrative.
The project also examines the boundary between visual realism and documentary credibility. Because the intended communication depended heavily on human identification and trust, the use of synthetic characters presented a strategic challenge extending far beyond image generation.
Click to explore the creative workflow Documentary credibility, organic imperfection, character continuity, generative production, editing and sound.
The Return is an independent AI-assisted documentary proof of concept developed during the exploratory stage of a production that was ultimately not realised.
The objective was to explore whether a fully generated production could sustain the restrained and observational language of documentary cinema without becoming visually artificial, excessively dramatic or conventionally promotional.
Client objective and realism strategy: The brief was not simply to produce a stylised AI proof of concept. The film was intended to resemble a documentary genuinely shot on location and to be as difficult as possible to distinguish from real documentary footage. This pursuit of photographic credibility shaped the generation, selection, direction and post-production of every scene.
The central communication challenge: The intended film was not merely required to look realistic. Its communication objective depended on emotional identification and trust. The audience needed to perceive the protagonist and surrounding community as credible human beings whose gestures, routines and relationships felt observed rather than artificially constructed.
Human presence was therefore central to the communication strategy, not simply to the visual style. The challenge was to determine whether synthetic characters could carry some of the emotional weight normally provided by real people, lived experience and authentic documentary observation.
Creative approach: The film follows an elderly man returning to the landscape and community with which he shares a deep personal connection. The emphasis is placed on ordinary gestures, rural atmosphere, natural pacing and the quiet relationship between character and place.
Organic imperfection as a realism tool: Traditional audiovisual production frequently attempts to eliminate unwanted errors, irregularities and inconsistencies. In generative filmmaking, however, an image that appears excessively clean, symmetrical or controlled can itself reveal its artificial origin.
Small irregularities in gesture, posture, timing, framing, environmental movement and incidental detail can make a scene feel observed rather than designed. In this context, a degree of organic imperfection can become desirable because it recalls the unpredictability and physical variation naturally present in documentary footage.
The objective was not to retain obvious generation failures, but to distinguish unwanted AI artefacts from imperfections that contributed to credibility. During scene generation, the prompts encouraged restrained performances, ordinary body language, natural environmental movement, non-heroic compositions and subtle variations associated with everyday life.
This approach continued during post-production. The generated images and clips were further shaped through shot selection, reframing, timing adjustments, visual treatment and traditional editing. Distracting artificial movements and generation errors were reduced or concealed, while useful irregularities were preserved when they helped the scenes feel less synthetic and more organically observed.
Character and visual continuity: Particular attention was given to preserving the protagonist’s identity, age, clothing, physical presence and emotional restraint across different locations and generated shots.
The surrounding architecture, landscape, lighting and time of day were also developed as parts of a coherent visual world rather than as isolated generated images.
Direction and sequence design: Each scene was conceived as part of a continuous cinematic progression. Framing, direction of movement, eyelines, scale and transitions were controlled to create the impression of a filmed documentary sequence.
Post-production and traditional editing: The AI-generated clips were treated as source material rather than as finished scenes. After generation, the film was constructed through a traditional non-linear editing workflow, selecting the strongest and most credible portions of each clip and combining them into a coherent sequence.
Editing, timing, reframing and shot selection were used to mitigate visual artefacts, continuity errors and unintended movements produced by the generative tools. This process also established the pace of the scenes and shaped the overall rhythm and progression of the narrative.
Image finishing and material realism: The generated material did not initially provide the degree of photographic texture and local specificity required by the brief. During post-production, controlled film grain was added and localised masks were used to treat selected areas independently. Particular attention was given to the texture and density of granite walls and loose stone, house façades, vegetation, and the behaviour of light across these surfaces.
Contrast, detail, colour, texture and local illumination were adjusted to reduce synthetic smoothness, correct inconsistent material qualities and create a more coherent impression of footage captured by a real camera in a real place.
Sound design and final mix: The physical environment was reconstructed through sound editing, sound design and mixing. Real recordings of wind, birds, footsteps, vegetation and distant village activity were selected from sound libraries, layered and carefully balanced to connect the generated scenes and give them greater realism, continuity and a credible sense of place.
Sound was not treated merely as accompaniment. It helped give weight and physical presence to gestures, environments and transitions that had originated as generated images, grounding the visual material in a recognisable acoustic reality.
Music: The soundtrack heard in this excerpt is also AI-generated, but it was not created by me. My contribution involved its editorial placement and integration with the environmental sounds and effects within the final mix.
Strategic and ethical risk: The project therefore operated deliberately at the boundary between synthetic documentary and visual simulation: one of its success criteria was the extent to which the generated scenes could be mistaken for footage captured on location. That ambition was central to the client’s communication strategy, but it also introduced a significant ethical risk. A synthetic character may appear visually convincing, but visual plausibility is not the same as lived experience or documentary truth.
If viewers believe they are watching real people, communities or testimony and later discover that the material was generated, the same realism intended to establish trust may instead weaken it. This risk becomes particularly relevant when the subject involves community, social impact, personal experience or a relationship of trust between an organisation and the people it hopes to reach.
The proof of concept therefore also became an evaluation of where AI-generated documentary language is appropriate. Generative production can be effective for visualising concepts, reconstructing situations or creating clearly presented documentary fiction. When credibility depends on authentic testimony, cultural knowledge or lived human experience, real filming or a transparent hybrid approach may remain the stronger creative and communication strategy.
This study demonstrates both the production potential and the strategic limitations of AI-generated filmmaking. Generative tools can create characters, environments and complete narrative sequences under consistent human direction, but technical realism alone cannot guarantee emotional authenticity or documentary credibility. Choosing between an AI-generated, hybrid or traditionally filmed approach must therefore depend on the communication objective and on the relationship of trust the film needs to establish with its audience.
Role: Concept development, creative direction, prompt design, generative visual development, character continuity, sequence design, traditional editing, visual post-production, mitigation of generative artefacts, sound editing, sound design and final mix.
Production: Independent proof of concept by Diogo Pessoa de Andrade / Creative Visuals.
WeTransact — The Mystery of the Giant Octopus
“Johan has been taken by a huge octopus! He was just dragged down the stairs!”
A corporate action-comedy created to celebrate WeTransact’s Microsoft Partner of the Year recognition. After CEO Johan Aussenac is apparently kidnapped by a giant inflatable octopus, his colleagues launch a rescue mission through the streets of Lisbon.
Generative AI was used as a production-rescue tool to reconstruct scenes that could not be filmed, repair continuity problems and create new performances involving recognisable employees wearing complex inflatable octopus costumes.
Click to explore the full production-rescue case study Missing scenes, continuity repair, AI performances, VFX and final post-production.
WeTransact — The Mystery of the Giant Octopus is a promotional comedy short created to celebrate WeTransact’s recognition as Microsoft Partner of the Year FY25 in the Start-Up category.
Instead of presenting the achievement through a conventional corporate announcement, the film transforms it into an absurd action-comedy. After WeTransact CEO Johan Aussenac is apparently kidnapped by a giant inflatable octopus, his colleagues launch a rescue mission through the streets of Lisbon, culminating in a final confrontation at Praça do Comércio.
The apparent monster ultimately has a less threatening purpose: ensuring that the Microsoft award reaches the people who earned it.
Production challenge: The original script was highly ambitious for a single-day shoot. The cast consisted of WeTransact employees rather than professional actors, and filming had to be coordinated around their normal work and client commitments.
The production also involved multiple Lisbon locations, live-action chase scenes and inflatable octopus costumes that depended on portable air pumps. With a limited shooting window and considerable pressure to complete the script, several important narrative moments could not be filmed, while other scenes were captured without every connecting action required for seamless continuity.
Because traditional reshoots were no longer practical, the missing material and some of the continuity problems identified during editing were resolved through a hybrid workflow combining generative AI, original live-action footage and traditional visual-effects techniques.
AI scene generation: Entirely new scenes were created to complete missing narrative moments while preserving continuity with the original cast, environments and comic tone. AI-generated versions of the performers were directed to carry out actions and deliver dialogue that had not been captured during the live-action shoot.
One particularly demanding aspect of this process was the need to recreate human performers inside inflatable octopus suits. This required maintaining consistency not only in facial identity, expression, direction of gaze and body language, but also in the design and physical behaviour of the costumes.
The shape, scale, inflation, silhouette and position of the tentacles had to remain visually coherent across different shots. The interaction between the performers’ bodies and the inflatable material also needed to feel believable so that the generated scenes could sit convincingly beside the original footage.
Continuity repair and narrative reconstruction: Because the production was following an ambitious script within a very restricted schedule, some continuity mismatches only became apparent during the editing process.
In several sequences, the live-action material did not contain all the connecting shots, character movements or visual transitions needed to maintain a clear and fluid progression. Changes in position, direction of movement, costume behaviour or character interaction risked interrupting the continuity of the scene.
AI was therefore used not simply to replace footage that had not been recorded, but also to create new transitional scenes and character moments that had never existed during the original production.
These additional shots helped correct continuity problems, bridge gaps between performances and locations, and preserve the internal logic of the narrative without requiring a traditional reshoot.
In some cases, entirely new character actions were generated specifically to establish cause and effect between two existing shots. This allowed the characters’ movements, reactions and interactions to remain visually understandable while preserving the rhythm and comic timing of the film.
Voice reconstruction and lip-sync: AI voice synthesis was used to recreate missing dialogue while preserving the tone, rhythm and accents of the original performances.
In one sequence, the performer preferred to use his own recorded voice. The dialogue was therefore captured separately during post-production and then carefully synchronised with the mouth movements of the AI-generated character.
VFX, tracking and compositing: The physical award presented another substantial visual challenge because it is made from transparent glass, with visible reflections, refractions and embedded logos.
AI-generated elements were combined with traditional rotoscoping, tracking, match moving and compositing to reconstruct the trophy, preserve its authentic branding and integrate it into shots that had not been captured during filming.
The generated octopus sequences also required additional compositing and visual supervision to match facial likeness, costume inflation, tentacle placement, scale, lighting and overall silhouette between the AI-created material and the live-action footage.
Additional generated elements were integrated directly into the filmed shots, while aerial images of Lisbon were combined with the original material to expand the visual scale of the chase.
Editing, motion graphics and sound: The final narrative was shaped through detailed human-led post-production. Dynamic editing, animated cartoon titles, custom motion graphics, music, layered sound effects and sound design were used to control the pace, strengthen the action and ensure that the comedic moments landed effectively.
Colour and visual continuity: The finished film combines footage recorded with professional cinema cameras, GoPro cameras, aerial material and AI-generated sequences. Colour correction, grading and compositing were used to match these different sources and create a coherent visual identity throughout the film.
This project demonstrates how AI-enhanced filmmaking can function not only as a creative tool, but also as a practical production-rescue and continuity solution. When real-world limitations prevented every planned scene and connecting action from being captured, generative technologies provided missing performances, transitions, environments and narrative elements.
The final result was only made possible by combining those tools with traditional directing, editing, rotoscoping, visual effects, sound design, motion graphics and colour finishing. AI supplied some of the missing pieces, but the film’s structure, continuity, timing, humour and final audiovisual language remained human-directed.
Role: Direction, editing, AI integration, generative scene development, continuity reconstruction, voice and lip-sync workflow, visual effects, tracking, match moving, rotoscoping, compositing, motion graphics, sound design, colour grading and final post-production.
Script: Julia do Prado / WeTransact
Production: Creative Visuals for WeTransact
Creative Visuals — Ideas that Talk and Walk
“Hey you, just look at this! I’m an idea, and I can talk and walk.”
An experimental hybrid short that brings a brand idea to life through a playful blue cat mascot walking through Lisbon. The character represents an abstract creative idea made visible, mobile and engaging.
Original live-action cinematography was combined with generative character development, AI voice synthesis, logo animation, sound design and traditional post-production.
Click to explore the creative and AI workflow Live-action filming, mascot development, AI voice, sound and brand integration.
Creative Visuals: Ideas that Talk and Walk is an experimental hybrid short that brings a brand idea to life through character-based storytelling.
Set on the streets of Lisbon, the piece introduces a playful blue cat mascot as a visual metaphor for creative ideas made visible, mobile and engaging.
The concept explores how a brand identity can move beyond a static logo or abstract message and become a living character inhabiting the real world. By combining local cinematography, generative character creation and post-production, the project presents an imaginative and cinematic way of expressing what Creative Visuals represents.
Creative approach: The original footage was captured in Lisbon and used as the foundation for the piece. The concept, visual direction, pacing, brand framing and final structure were developed through a human-led filmmaking process, with careful attention to atmosphere, humour and visual identity.
AI workflow:Artificial intelligence was used to iteratively develop the blue cat mascot, integrate the character into the filmed environment and support the creation of additional visual elements such as the aviator goggles and other design refinements.
AI voice synthesis was also used to generate the character’s English voice performance, helping define its playful and distinctive personality.
Props, sound and finishing: The piece incorporates a rare vintage Yashica-44 camera as part of the narrative and visual identity.
The final result was shaped through traditional post-production, including editing, sound design, layered audio effects, logo animation and colour grading.
This short is an example of hybrid filmmaking: AI accelerated the creation and integration of character-based assets, while the creative intention, cinematic treatment, timing, soundscape and brand expression remained fully guided by human direction.
Role: Concept development, live-action filming, creative direction, AI character generation and integration, voice development, editing, sound design, logo animation, colour grading and post-production.
Ramma — O Inferno Arde em Mim
“O inferno arde em mim.”
A dark, cinematic music video exploring grief, memory and inner torment, filmed entirely at night at the historic Palace of the Marquis of Pombal in Oeiras.
The production was developed primarily through traditional cinematography, practical lighting and post-production. Generative AI was used only for the climactic chapel sequence, where the projected angel wings ignite and additional flames appear around the altar and within selected elements of the chapel.
The visual effects had to remain coherent throughout a forward tracking shot following the artist towards the altar, preserving the camera movement, changing perspective and interaction of the simulated firelight with the surrounding architecture.
Click to explore the hybrid cinematography and AI VFX workflow Night cinematography, practical projection, moving-camera AI effects, generative fire, compositing and colour finishing.
Ramma — O Inferno Arde em Mim is a dark, cinematic music video exploring grief, memory and inner torment.
The video was filmed entirely at night at the historic Palace of the Marquis of Pombal in Oeiras. Its underground spaces, formal gardens and baroque chapel were used as distinct narrative environments reflecting different emotional stages of the song.
Traditional production: The project was conceived primarily as a conventional live-action music video. The performances, locations, camera movements, lighting and visual atmosphere were created during a multi-day shoot using traditional filmmaking methods.
Low-light cinematography and controlled practical lighting were used to preserve the architecture of the palace while creating a dark, gothic visual language appropriate to the music.
The chapel challenge: The intended climax required the projected angel wings to transform into fire, with additional flames appearing around the altar and within selected architectural elements of the chapel.
Using real flames inside this protected historic space was neither safe nor permitted. The scene also involved a forward tracking shot, moving from the rear of the chapel towards the altar while following the artist. Any added visual effect therefore had to adapt continuously to the changing camera position, perspective and composition.
During filming, a 600-watt light projector with optical modifiers was used to cast static blue angel wings onto the altar. This practical projection established the composition, scale, light direction and visual foundation for the transformation developed later in post-production.
AI-assisted visual effects: Generative AI was used exclusively for this climactic sequence. The static projected wings were given fluid movement and transformed into burning wings, while additional flames were generated around the altar and within selected elements of the surrounding chapel.
Because the camera travels forward behind the artist, the generated wings and flames could not behave like a static overlay. Their position, scale, perspective and relationship with the architecture had to remain coherent as the camera moved closer to the altar.
The simulated firelight also had to interact convincingly with glass, metal, statues, walls and other architectural surfaces throughout the movement. Reflections, illumination and changes in intensity were developed to support the impression that the flames belonged within the filmed environment.
The generated material was not used as a complete replacement for the original shot. It was developed from the filmed composition and combined with the practical wing projection, the artist’s live-action performance, the existing chapel lighting and the original travelling camera movement.
Artist approval: The artist was initially opposed to the use of artificial intelligence in the music video. A test of the chapel transformation was therefore produced before the sequence was included in the final edit.
After seeing how the animated wings, generated fire and environmental light interaction could be integrated with the original cinematography, he approved the sequence for the completed music video.
Post-production and finishing: The generated elements were selected, refined and integrated through a traditional post-production workflow involving editing, visual effects, compositing and colour grading.
Particular attention was given to preserving the forward camera movement, the artist’s silhouette and performance, the chapel architecture, the changing perspective, the original lighting direction and the visual continuity between the practical and generated elements.
The result demonstrates a selective hybrid workflow in which generative AI resolves a specific physical and production limitation without replacing the live-action foundation of the project.
In this case, AI functioned as an additional visual-effects tool, allowing an otherwise impractical scene to be created inside a protected location while retaining the performance, cinematography, practical projection and moving-camera shot captured during production.
Role: Co-concept development, direction, cinematography, editing, post-production, AI-assisted visual effects, compositing and colour grading.
Concept: Ramma and Diogo Pessoa de Andrade
Production: Creative Visuals
The Dilemma — Hey Zé, Come Here
“Sometimes words are not enough. And expectations... well, those can surpass any reaction.”
An experimental AI-animated comedy short about the unrealistic expectations people sometimes project onto their partners.
What begins as a domestic disagreement becomes an absurd comic-book-style escape, using character animation, synthetic voices, editing and sound design to construct the performances and final punchline.
Click to explore the creative and AI workflow Character animation, synthetic voices, editing, sound design and comedic timing.
The Dilemmais an experimental AI-animated comedy short inspired by a familiar relationship dynamic: the unrealistic expectations people can project onto their partners, even when those expectations are never clearly expressed.
What begins as a typical domestic disagreement gradually turns into an absurd escape worthy of a comic-book hero. By exaggerating the breakdown in communication, the film transforms an everyday situation into a short piece of visual comedy.
Creative approach: The concept, script, characters, dialogue and directorial vision were developed through a traditional human-led creative process.
The timing of the performances, escalation of the argument and final punchline were shaped through editing, sound and careful control of rhythm.
AI workflow: AI-assisted character animation was used to bring movement, facial expression and physical performance to the original static characters.
AI voice synthesis was also used to create the dialogue performances and explore the intended comedic tone.
Editing, sound and finishing: The generated material was selected, organised and refined through traditional post-production.
Editing, English subtitling, music, sound effects and sound design were combined to build the atmosphere and ensure that the comic timing remained precise.
This project illustrates a hybrid filmmaking workflow in which artificial intelligence functions as a production tool rather than the author of the work.
AI reduced the time and resources required for character animation and voice production, while the concept, narrative decisions, pacing, emotional intention and final audiovisual construction remained human-directed.
Role: Concept, scriptwriting, character creation, creative direction, AI generation and animation, voice direction, editing, English subtitles, sound design and post-production.
Karl Marx Sells an iPhone
“Don’t buy a phone. Buy the crystallised unpaid labour of the proletariat.”
A short political satire in which Karl Marx appears to promote an iPhone through the language of exploitation, alienation and capitalist consumption.
The piece began with an AI-generated black-and-white image of Marx holding a modern smartphone. The static image was animated and combined with an AI-generated Russian voice performance.
After the AI generation stage, the final piece was constructed through human editing, voice synchronisation, sound design, audio post-production, colour work and old-film effects.
Originally created as a playful provocation for a communist friend, the experiment also demonstrates how generative tools can give movement and voice to historical figures for satire, period films, documentaries and other creative work.
Click to explore the historical-character and AI workflowHistorical-image recreation, character animation, synthetic voice, editing, sound design and vintage film finishing.
Karl Marx Sells an iPhoneis a short AI-generated satire that places one of the most influential critics of capitalism in the role of a commercial spokesperson for a modern consumer product.
Holding a smartphone towards the viewer, Marx describes the device not as a telephone, but as crystallised unpaid labour, a polished object of alienation and a form of oppression that the consumer can take home.
The humour comes from the collision between Marxist language and the visual conventions of commercial advertising. A historical figure associated with the critique of commodities is transformed into someone apparently attempting to sell one.
Origin of the project:The video was initially created as an inside joke for a communist friend with whom I regularly discuss politics.
The intention was not to produce a serious political statement, but a playful provocation based on the contradiction between Marx’s ideas and the symbolism of a premium smartphone as an object of consumption, technology and social status.
Concept and script:I wrote the original satirical text by adapting concepts associated with Marxist theory to the language of a product advertisement.
“Guys, don’t buy a phone. Buy the crystallised unpaid labour of the proletariat, a shiny piece of alienation. Take your oppression home.”
Although Karl Marx was German, the script was translated into Russian as a deliberate satirical and aesthetic choice. The language evokes the later Soviet and communist imagery commonly associated with Marx, while reinforcing the exaggerated tone of the imaginary vintage advertisement.
AI-generated source image:The project began with an AI-generated static image of Karl Marx holding a modern silver smartphone and presenting it towards the viewer.
The composition was designed like a direct-to-camera product advertisement: Marx holds the telephone in one hand and points towards it with the other, addressing the viewer as though demonstrating the qualities of the product.
Historical-character animation:The static portrait was animated using generative AI to create facial movement, speech, hand gestures and small changes in posture.
Particular attention was given to preserving the identity of the character, the proportions of the face and beard, the position of the telephone and the relationship between the pointing hand and the product.
The objective was not to create perfectly contemporary or polished movement. The slightly abrupt gestures and imperfect motion became part of the intended vintage aesthetic, resembling footage recorded with an early low-frame-rate camera.
Voice generation:An AI-generated male voice was used to perform the translated Russian script.
The delivery was directed to remain calm and discursive while retaining an ironic advertising tone. The contrast between the serious voice and the absurd sales pitch strengthens the satirical effect.
Editing, sound and finishing:After generating the animated material and voice with AI, I assembled and refined the final piece through traditional post-production.
The generated voice was synchronised with the animated image during editing, and the soundtrack was developed through sound design and audio post-production to integrate the speech convincingly with the visual rhythm of the scene.
The final sound treatment was therefore not simply the direct output of the voice-generation tool. The voice, timing and supporting sound elements were combined and adjusted during post-production.
Vintage visual treatment:After the AI-generated animation was completed, the material was edited and treated to resemble an early black-and-white television commercial or an advertisement from the first decades of filmed media.
Colour correction and contrast shaping were used to create the final monochrome appearance. Film grain, flicker, vignette, image instability, scratches and other old-film effects were then added in post-production.
The video is presented in a 16:9 aspect ratio, while the monochrome image, film grain, flicker, vignette, image instability and intentionally jerky movement create the impression of aged archival footage.
These visual imperfections help integrate the generated animation and give the piece the appearance of an archival commercial discovered from an alternative historical timeline.
The finished result therefore combines AI-generated imagery, animation and voice with human editing, voice synchronisation, sound design, audio finishing, colour work and visual post-production.
Beyond the joke:Although this particular video was created as a humorous political provocation, the underlying process has wider creative applications.
Generative animation can give movement and voice to historical figures reconstructed from photographs, portraits and other archival visual material. This can support period films, historical documentaries, museum installations, educational projects and fictional works.
The same approach can be used to visualise historical testimony, create stylised reconstructions or explore how people from the past might be represented within a contemporary audiovisual language.
Such uses require careful consideration of historical accuracy, context, transparency and the distinction between documented material and creative reconstruction.
This short demonstrates the process on a deliberately absurd scale: artificial intelligence transforms a static representation of a nineteenth-century thinker into the presenter of an imaginary vintage technology advertisement.
Role:Concept development, scriptwriting, translation workflow, AI image generation, historical-character animation, voice development, editing, voice synchronisation, sound design, audio post-production, colour grading, vintage film effects and final post-production.
AI-Assisted Music and Cinematic Sound
These projects form a complementary strand of the same filmmaking practice, using original poetry, concepts and audiovisual direction to develop voices, rhythm, atmosphere and cinematic soundworlds through generative music, arrangement, micro-editing and sound design.
O puto e o velho — From Poem to Cinematic Soundworld
“A poem that travelled through more than thirty years before finding a voice and a sound.”
O puto e o velho (“The Boy and the Old Man”) grew out of an original poem I wrote in the 1990s. The work explores the struggle of living, the passage between generations and the exchange between the innocence of someone beginning life and the awareness of someone approaching its end.
Decades later, the poem was transformed into a musical and audiovisual piece through a human-led process combining AI-assisted music development with arrangement, selection, restructuring, editing, mixing and sound design.
The project demonstrates how generative music can provide raw material for developing a specific cinematic soundworld shaped around an existing story, emotional progression and visual narrative.
Click to explore the music and sound development workflowOriginal poetry, AI-assisted music development, arrangement, editing, mixing and cinematic sound design.
O puto e o velho began as a poem written in the 1990s, long before the musical and generative technologies used to give it its present form existed.
The original poem: The narrative follows an encounter between a boy and an old man. Through images of bread, stone, strength, pain and death, it reflects on the struggle of living and on experience passing from one generation to the next.
The text moves from the purity, passion and dreams associated with the beginning of life towards the hardness and accumulated awareness of someone who has travelled an entire lifetime. Its conclusion accepts death not simply as an ending, but as an essential part of a cycle in which life continues through those who remain.
AI-assisted music development: Generative music tools were used to explore different voices, arrangements, atmospheres and musical interpretations of the original Portuguese lyrics.
The finished work was not taken directly from a single generation. Numerous results were reviewed and compared, after which selected material was restructured and refined according to the narrative progression, emotional rhythm and intended cinematic atmosphere.
Arrangement and reconstruction: Sections were reorganised and adjusted during editing. A brief vocal passage inspired by Cante Alentejano was selected from a separate generation attempt and integrated into the final structure.
The entrance and exit of this passage were aligned with the musical pulse, while transitions, pauses and changes in intensity were refined to preserve the emotional continuity of the piece.
Mixing and sound design: The selected musical material was assembled, balanced and treated through a traditional post-production workflow. Editing, mixing, atmospheric elements and sound design were used to transform the generated sources into a coherent finished work.
Application to film: This workflow can support the development of musical concepts, mood pieces and cinematic soundworlds shaped around the rhythm, emotional direction and narrative requirements of a film.
It can be particularly useful for poetic films, short films, documentaries, visual essays, concept pieces and independent productions requiring an original relationship between words, images and sound.
Role: Original poem and lyrics, creative concept, AI-assisted music development, arrangement, selection, editing, mixing, sound design, visual concept and final audiovisual construction.
AI-assisted music creation: Suno.
Release and reuse: The finished work is available under the Creative Commons Attribution 4.0 International licence and may be reused, adapted and included in commercial projects with appropriate attribution.
Atrevo-me — Sensual Poetry and Cinematic Sound
“Desire shaped by voice, breath, silence and sound.”
Atrevo-me (“I Dare”) is an intimate, sensual and erotic musical poem. Its lyrical progression begins with imagination and anticipation, moves through touch, desire and physical climax, and ultimately passes from the body towards spirit, absence and reunion.
The rhythm is led less by a rigid beat than by the weight of the words, the cadence of the male voice, the female breaths and the silences between them. A recurring tenor saxophone enters from the opening section and returns throughout the piece, creating transitions, tension and release.
The final version was shaped through iterative generation, selection, restructuring and multitrack editing. Words, breaths, pauses, vocal textures and instrumental fragments were organised around the dramatic progression of the poem rather than accepted as an untouched generated performance.
Click to explore the lyrical, musical and cinematic sound workflowErotic poetry, vocal direction, micro-editing, generative music, sound design and AI-assisted poster development.
Atrevo-me began with an original Portuguese text built as a sensual and erotic crescendo. Images of lips, hair, skin, breath and physical pleasure gradually intensify until the body reaches its climax. The final movement then changes perspective, moving “from matter to spirit” as intimacy dissolves into time, absence and rediscovery.
Rhythm led by language: The piece does not depend on a continuous or mathematically rigid beat. Its pulse emerges from phrasing, the weight of individual syllables, suspended pauses, breathing and silence. The delivery frequently approaches spoken word and intimate declamation, allowing each interruption or hesitation to generate dramatic and erotic tension.
Vocal direction: The male performance was conceived as an extremely low, mature bass-baritone: gravelly, smoky, restrained and close to the listener. The female presence remains wordless and is expressed through deep breathing, humming, sighs and a brief vocal cry at the height of the piece.
The saxophone as a dramatic voice: The tenor saxophone is not reserved for a conventional solo or ending. It appears near the beginning and returns between lyrical sections, creating a sensual dialogue with the voices. Its recurring phrases act as transitions, extend the emotional space and give the music a cinematic rather than conventionally song-driven structure.
AI as source material: Generative music was used to explore vocal textures, instrumental phrases, silences, dynamics and spatial treatments. The generated material was approached as a bank of possible performances and sound fragments, rather than as an automatically finished composition.
Micro-editing and multitrack construction: Individual words, breaths, pauses, vocal moments and instrumental passages were selected, separated and repositioned according to expressive cadence rather than being locked to a rigid rhythmic grid. This process gave human control over timing, escalation and the relationship between the poem and the instrumental development.
Architecture of the climax: The increasing physicality of the lyrics is mirrored by the vocal and instrumental intensity. The central erotic climax culminates in humming and a short vocal cry, followed by an extended musical space. The final verses then withdraw from the body towards the spiritual, before the piece closes with deep breathing, a fading hum and a final sigh.
Editing and sound treatment: Transitions, level relationships, tonal balance, pauses and degrees of reverberation were refined after generation. The contrast between dry vocal proximity and wider, darker resonant spaces helps transform the generated sources into a coherent and deliberately shaped soundworld.
Visual development: The accompanying image was developed through an iterative AI-assisted process guided by a sensual cinematic poster concept. Character age, facial position, gaze, expression, darkness, typography and the relationship between the two profiles were progressively refined before the final audiovisual edit.
From music to cinematic soundworld: The project reflects a filmmaker’s approach to music: sound is organised around dramatic progression, atmosphere, rhythm and its possible relationship with images. This process can support the creation of original sonic universes, musical concepts and soundtrack material for films, poetic works, visual essays and other narrative audiovisual projects.
Role: Original lyrics and poetry, creative concept, lyrical and dramatic structure, vocal and instrumental direction, generative development, selection, micro-editing, multitrack reconstruction, sound treatment, visual concept, AI-assisted image direction, poster development and final audiovisual construction.
AI-assisted music creation: Suno.
Release and reuse: The finished track may be used and adapted free of charge, including in commercial projects, provided that the original lyrics and creative direction are credited to Diogo Pessoa de Andrade. The work will not be registered with Content ID.
Los Pura Pose — Posa para mim
“Claquete. Centro. Esquerda. Centro. Direita. Centro. Cima. Centro. Baixo.”
An experimental music video born from a real production created to record human facial expressions for the training of an artificial-intelligence system.
During the original five-day shoot, participants were filmed performing different predefined expressions while an audio guide instructed them to turn their faces towards the centre, left, right, up and down. Those functional recording commands later became the lyrics and rhythmic basis of a reggaeton-influenced track created with generative AI.
The song and music video were created in a single day. The finished video combines the original instructional recording, AI-generated music, Creative Commons stock footage and rhythm-driven human editing.
Click to explore the real-world origin and creative workflow Facial-expression capture, original field audio, AI-generated music, Creative Commons footage and rhythm-driven editing.
Los Pura Pose — Posa para mim is an experimental AI-assisted music video inspired by a real audiovisual data-capture production.
I was hired to assemble and coordinate a production team responsible for filming people performing different facial expressions. According to the information provided to us, the resulting recordings would be used to help train an artificial-intelligence system to recognise expressions and emotional states.
The original production: The recording process took place over five days inside an apartment adapted to contain four separate filming stations.
Three stations were located in different interior rooms and one was positioned outside on a balcony. Each location had a precisely defined level of illumination measured in lux at the participant’s face.
The stations covered lighting conditions ranging from approximately 10 lux in the darkest setup to more than 4,000 lux in exterior daylight. This allowed the same expressions and facial movements to be recorded under very different lighting conditions.
Participants were filmed at four predefined distances from the camera, marked physically on the floor. At each distance and lighting station, they performed eight expressions selected from a predefined group that included neutral, happy, angry, excited, shocked or fearful, affectionate, disgusted, disappointed and other emotional states.
For every expression, the participant had to turn their face through a precise sequence of directions so that it could be recorded from multiple angles.
The instructional recording: To maintain consistent timing throughout the sessions, an audio guide was played during every take.
The recording began by calling for the slate, which displayed the participant’s assigned identification number and the identification number of the expression being performed.
A timed voice then instructed the participant to move through the following sequence:
“Claquete. Centro. Esquerda. Centro. Direita. Centro. Cima. Centro. Baixo. Centro.”
The same sequence was repeated across the different expressions, camera distances and lighting stations. After five days of production, the rhythm and repetition of those commands had become extremely familiar.
From technical instruction to song: After returning home, I decided to transform the functional recording process into a playful musical experiment.
The exact directional commands used during filming became the basis of the lyrics. Rather than writing a conventional narrative song, I retained the mechanical language of the production:
“Claquete, centro, esquerda, centro, direita, centro, cima, centro, baixo.”
Generative AI was then used to transform those words into a reggaeton-influenced track. The result converts a practical set of instructions into a repetitive musical command that invites the viewer to move, pose or follow the directions.
The opening of the song includes a mix of the original voice recording used during the real filming sessions and the newly generated musical track. After this introduction, the piece transitions fully into the AI-generated music.
This connection between the original production audio and the finished song preserves the real-world origin of the project rather than merely imitating the language of a fictional photo shoot.
Music-video construction: The visual component was created using stock footage made available under Creative Commons licences.
The selected images feature different people, ages, appearances and expressions, visually echoing the diversity and facial-performance focus of the original recording project.
The footage was selected, reorganised and edited to respond to the musical instructions. Changes of expression, gaze, pose, direction and framing were synchronised with the spoken commands, percussion and rhythmic accents of the track.
The performers seen in the music video are therefore not AI-generated characters. The AI contribution is primarily musical, while the visual structure was created through the selection and transformation of existing Creative Commons footage in post-production.
Editing and rhythm: The repetition of “centre”, “left”, “right”, “up” and “down” provided a simple but strict framework for the edit.
Individual shots were timed and rearranged so that facial movements, poses and visual changes interact with the instructions in the song. The editing transforms otherwise unrelated stock footage into a coherent visual performance with its own rhythm, escalation and comic energy.
Created in one day: Both the song and the completed music video were developed in a single day as a spontaneous creative response to the production experience.
The project demonstrates how material originating in a highly controlled and technical workflow can be reinterpreted through generative music, archival selection and human editing.
Rather than using AI to reproduce the original assignment, the piece reverses its logic: a process designed to collect human expressions for an algorithm becomes the inspiration for a human-directed audiovisual work.
Role: Original production-team coordination, creative concept, lyrics, AI music development, Creative Commons footage research and selection, editing, audiovisual synchronisation and final post-production.
Tools Don’t Make the Craftsman
An old proverb says that “the cowl does not make the monk.” In the same way, access to sophisticated tools does not automatically create knowledge, judgement or craftsmanship.
A camera does not make someone a filmmaker, and neither does access to an AI model. Tools can expand what is technically possible, but meaningful work still depends on ideas, experience, visual judgement and the ability to make deliberate creative decisions.
Generating an image, a voice or a video clip is only one part of the process. The greater challenge is creating material that feels intentional, coherent and connected to the story, brand or message behind the project.
My role is to guide the process from the initial concept and visual language through generation, selection, editing, sound, colour and final delivery, so that the technology serves the project rather than defining it.
What AI Can Add to a Film or Video Production
- Concept development and visual research: exploring characters, environments, moods and visual directions before production begins.
- Pre-visualisation: creating style frames, storyboards, visual treatments and early versions of scenes or sequences.
- Generative video and imagery: producing original visual material for films, advertising, branded content, documentaries, music videos and online campaigns.
- Complex or inaccessible scenes: visualising historical settings, imaginary worlds, conceptual imagery or shots limited by budget, logistics, access, rights or available archive material.
- Hybrid live-action productions: combining traditionally filmed footage with AI-generated or AI-transformed elements.
- Continuity and production repair: creating missing connecting shots, reconstructing scenes and resolving visual or narrative continuity problems discovered during editing.
- Image transformation and extension: modifying environments, extending shots, developing transitions or creating alternative visual treatments from existing material.
- AI-assisted post-production: supporting editing, compositing, image restoration, cleanup, transcription, subtitles and versioning.
- Audio, voice and music: supporting dialogue restoration, voice generation, multilingual versions, sound design and the development of original musical ideas.
Live Action and Generative AI
AI filmmaking does not need to exist separately from traditional production. Some of the most interesting possibilities come from combining real cinematography, locations, performers and practical elements with generative imagery and AI-assisted post-production.
A production may begin with original live-action footage and use AI to extend a location, transform a visual element, create an additional shot, develop a transition or introduce imagery that could not be captured during filming.
The opposite approach is also possible: AI-generated material can provide the starting point for a sequence that is later refined through editing, compositing, sound design, colour correction and other traditional post-production techniques.
This hybrid approach makes it possible to retain the authenticity and detail of filmed material while expanding the visual possibilities available to the project.
AI-Enhanced Audio, Voice and Sound
Sound is an essential part of storytelling. Alongside generative imagery and video, AI can support different stages of audio production, from improving recorded dialogue to developing voices, music and soundscapes for a finished film.
Depending on the needs of the project, AI-assisted audio workflows may include:
- Dialogue cleanup and restoration: reducing noise, improving clarity and recovering recordings made in challenging conditions.
- Transcription and subtitles: accelerating the preparation of transcripts, captions and translated subtitle versions.
- Voice generation and multilingual adaptation: creating narration, temporary voice tracks, alternative language versions or carefully authorised voice reproductions.
- Music development:exploring musical ideas, moods and compositions created or adapted to the rhythm and emotional direction of a film.
- Sound design and atmosphere: developing ambience, textures and supporting sonic elements that strengthen the visual world of the project.
As with image generation, these tools are used selectively and remain part of a human-led creative process. Voice rights, performer consent, music usage and the intended distribution of the project are considered before the final material is produced or released.
A Filmmaker-Led AI Workflow
Creative direction
Every project begins by defining its purpose, audience, visual language and intended emotional response. AI is introduced only where it can genuinely strengthen the idea.
Choosing the right tools
I do not restrict the process to one platform or model. Different tools are tested and selected according to the required visual style, movement, consistency, resolution and type of production.
As generative technology continues to evolve, the specific tools may change. The creative objective, rather than the platform, remains at the centre of the workflow.
Generation, selection and refinement
Generating material is only one stage of the process. Outputs are reviewed, compared, selected and refined. When necessary, they are regenerated, transformed or combined with other visual elements until they support the intended direction.
Editing and post-production
The selected material is integrated into a professional post-production workflow that may include editing, compositing, colour correction, sound design, music, voice, titles, subtitles and final mastering.
Consistency and quality control
Particular attention is given to continuity, character consistency, composition, movement, lighting, rhythm and the visual details that distinguish intentional filmmaking from generic AI-generated content.
Where AI Filmmaking Can Be Useful
- Commercials and branded films
- Corporate and institutional videos
- Documentaries and historical reconstructions
- Music videos and visualisers
- Short films and fictional sequences
- Concept films and mood pieces
- Product visualisation
- Social media and online campaigns
- Pitch films, treatments and pre-visualisation
- Hybrid productions combining live action and generative imagery
- Production rescue and continuity reconstruction
- Multilingual voice and video adaptations
Responsible and Considered Use of AI
The use of AI in professional production should be deliberate and appropriate to the project. Depending on the content and intended distribution, this may involve considering copyright, source material, likeness and voice rights, privacy, client confidentiality and transparency with the audience.
When AI is used with client material, sensitive information or unpublished creative work, the workflow should also take into account how data is processed and stored by the platforms involved.
I avoid treating AI as a shortcut for replacing creative judgement or professional craftsmanship. The objective is to use these technologies in a way that protects the identity of the project and maintains the quality, credibility and human perspective expected by the client.
AI-assisted music is treated as one possible creative workflow, not as a universal replacement for composers or performers. When a project calls for the authorship, interpretation or cultural knowledge of professional musicians, collaboration remains essential.
A Wider Reflection on AI and Filmmaking
AI is opening creative possibilities that, until very recently, were inaccessible to independent filmmakers. Images, worlds and sequences that once required large crews, substantial budgets and complex production infrastructures can increasingly be explored by a single creator.
That potential is extraordinary — but so is the disruption that comes with it. The same technologies that democratise production also reduce costs, accelerate workflows and may diminish the need for many of the people traditionally involved in filmmaking and audiovisual production.
As increasingly sophisticated tools become capable not only of generating images, video, voices and music, but also of assisting with writing, concept development and creative decisions, the impact may extend far beyond individual production tasks. It raises difficult questions about how creative work will be valued, priced and sustained in a market where technically complex content can be produced with fewer people and at dramatically lower cost.
I see AI neither as something to fear nor as something to celebrate uncritically. It is a powerful creative tool capable of removing barriers and expanding what an individual filmmaker can imagine and produce, while simultaneously contributing to a profound transformation of the industry around us.
Learning to work with these technologies therefore also means remaining conscious of their wider consequences for filmmakers, crews, artists and other professionals whose work has traditionally formed part of the audiovisual production process.
Related Film and Video Services
- Film Director & Filmmaker
- Cinematographer / Director of Photography
- Videographer Services
- Film Production Services in Portugal
- Advertising & Branded Content
- Films & Documentaries
- Client Reviews
Let’s explore how AI can expand the creative possibilities of your next project.
Tell me about itFilmmaker, Producer, Director & Cinematographer (DoP) / Videographer
Based in Lisbon, working across Portugal and internationally.